Fundamental of Atmospheric Modeling - Jacobson

This page intentionally left blank
Fundamentals of Atmospheric Modeling

Second Edition
This well-received and comprehensive textbook on atmospheric pro-
cesses and numerical methods has been thoroughly revised and updated.
The new edition includes a wide range of new numerical techniques
for solving problems in areas such as cloud microphysics, ocean
atmosphere exchange processes, and atmospheric radiative proper-
ties. It also contains improved descriptions of atmospheric physics,
dynamics, radiation, aerosol, and cloud processes. Numerous exam-
ples and problems are included, with answers available to lecturers at
http://www.cambridge.org/0521548659
Fundamentals of Atmospheric Modeling is essential reading for
researchers and advanced students of atmospheric science, meteorology,
and environmental science.
Mark Z. Jacobson is an associate professor of civil and environ-
mental engineering at Stanford University. Goals of his research are to
improve our understanding of physical, chemical, and dynamical pro-
cesses in the atmosphere through numerical modeling and to improve
the simulation of air pollution, weather, and climate. He is the author of
two textbooks: Fundamentals of Atmospheric Modeling and Atmospheric
Pollution: History, Science, and Regulation.
Fundamentals of
Atmospheric Modeling
Second Edition
MARK Z. JACOBSON
Stanford University
caxniioci uxiviisiry iiiss
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore, So Paulo
Cambridge University Press
The Edinburgh Building, Cambridge cn: :iu, UK
First published in print format
isnx-:+ ,-c-s::-+,c-,
isnx-:+ ,-c-s::-sos-o
isnx-:+ ,-c-s::-::::s-c
Cambridge University Press 2005
2005
Information on this title: www.cambridg e.org /9780521839709
This book is in copyright. Subject to statutory exception and to the provision of
relevant collective licensing agreements, no reproduction of any part may take place
without the written permission of Cambridge University Press.
isnx-:c c-s::-::::s-c
isnx-:c c-s::-+,c-x
isnx-:c c-s::-sos-,
Cambridge University Press has no responsibility for the persistence or accuracy of
uiis for external or third-party internet websites referred to in this book, and does not
guarantee that any content on such websites is, or will remain, accurate or appropriate.
Published in the United States of America by Cambridge University Press, New York
www.cambridge.org
hardback
paperback
paperback
eBook (EBL)
eBook (EBL)
hardback
To Dionna and Daniel
Contents
Preface page xiii
Acknowledgments xiv
1 Introduction 1
1.1 Brief history of meteorological sciences 1
1.2 Brief history of air-pollution science 5
1.3 The merging of air-pollution and meteorological sciences 6
1.4 Weather, climate, and air pollution 6
1.5 Scales of motion 8
1.6 Atmospheric processes 8
2 Atmospheric structure, composition, and thermodynamics 12
2.1 Pressure, density, and composition 12
2.2 Temperature structure 18
2.3 Equation of state 28
2.4 Changes of pressure with altitude 34
2.5 Water in the atmosphere 37
2.6 First law of thermodynamics 47
2.7 Summary 57
2.8 Problems 58
2.9 Computer programming practice 60
3 The continuity and thermodynamic energy equations 61
3.1 Denitions 61
3.2 Continuity equations 65
3.3 Expanded continuity equations 68
3.4 Thermodynamic energy equation 78
3.5 Summary 80
3.6 Problems 80
4 The momentum equation in Cartesian and spherical
coordinates 82
4.1 Horizontal coordinate systems 82
4.2 Newtons second law of motion 87
4.3 Applications of the momentum equation 111
4.4 Summary 135
vii
Contents
4.5 Problems 136
5 Vertical-coordinate conversions 138
5.1 Hydrostatic and nonhydrostatic models 138
5.2 Altitude coordinate 143
5.3 Pressure coordinate 143
5.4 Sigma-pressure coordinate 151
5.5 Sigma-altitude coordinate 160
5.6 Summary 167
5.7 Problems 167
6 Numerical solutions to partial differential equations 169
6.1 Ordinary and partial differential equations 169
6.2 Operator splitting 170
6.3 Advectiondiffusion equations 171
6.4 Finite-difference approximations 172
6.5 Series expansion methods 192
6.6 Finite-volume methods 199
6.7 Advection schemes used in air-quality models 199
6.8 Summary 202
6.9 Problems 202
7 Finite-differencing the equations of atmospheric dynamics 204
7.1 Vertical model grid 204
7.2 The continuity equation for air 208
7.3 The species continuity equation 211
7.4 The thermodynamic energy equation 213
7.5 The horizontal momentum equations 214
7.6 The hydrostatic equation 221
7.7 Order of calculations 222
7.8 Time-stepping schemes 222
7.9 Summary 224
7.10 Problems 224
7.12 Modeling project 225
8 Boundary-layer and surface processes 228
8.1 Turbulent uxes of momentum, energy,
and moisture 228
8.2 Friction wind speed 230
8.3 Surface roughness lengths 231
8.4 Parameterizations of kinematic turbulent uxes 235
8.5 Eddy diffusion above the surface layer 250
8.6 Ground surface temperature and soil moisture 254
viii
Contents
8.7 Summary 271
8.8 Problems 271
9 Radiative energy transfer 273
9.1 Energy transfer processes 273
9.2 Electromagnetic spectrum 275
9.3 Light processes 283
9.4 Absorption and scattering by gases and particles 290
9.5 Visibility 313
9.6 Optical depth 316
9.7 Solar zenith angle 317
9.8 The radiative transfer equation 320
9.9 Summary 334
9.10 Problems 334
10 Gas-phase species, chemical reactions, and reaction rates 336
10.1 Atmospheric gases and their molecular structures 336
10.2 Chemical reactions and photoprocesses 342
10.3 Reaction rates 344
10.4 Reaction rate coefcients 346
10.5 Sets of reactions 351
10.6 Stiff systems 353
10.7 Summary 355
10.8 Problems 355
11 Urban, free-tropospheric, and stratospheric chemistry 357
11.1 Free-tropospheric photochemistry 357
11.2 Urban photochemistry 375
11.3 Stratospheric photochemistry 393
11.4 Summary 415
11.5 Problems 416
12 Methods of solving chemical ordinary differential equations 418
12.1 Characteristics of chemical ODEs 418
12.2 Analytical solutions to ODEs 421
12.3 Taylor series solution to ODEs 421
12.4 Forward Euler solution to ODEs 422
12.5 Backward Euler solution to ODEs 424
12.6 Simple exponential and quasi-steady-state
solutions to ODEs 426
12.7 Multistep implicitexplicit (MIE) solution to ODEs 427
12.8 Gears solution to ODEs 432
12.9 Family solution to ODEs 439
ix
Contents
12.10 Summary 442
12.11 Problems 442
12.13 Modeling project 444
13 Particle components, size distributions, and size structures 446
13.1 Introduction to particles 446
13.2 Aerosol, fog, and cloud composition 447
13.3 Discrete size distributions 449
13.4 Continuous size distributions 454
13.5 Evolution of size distributions over time 462
13.6 Summary 467
13.7 Problems 468
14 Aerosol emission and nucleation 470
14.1 Aerosol emission 470
14.2 Nucleation 484
14.3 Summary 492
14.4 Problems 493
15 Coagulation 494
15.1 Implicit coagulation 494
15.2 Semiimplicit Coagulation 496
15.3 Comparison with analytical solutions 498
15.4 Coagulation among multiple particle distributions 500
15.5 Particle ow regimes 505
15.6 Coagulation kernel 508
15.7 Summary 522
15.8 Problems 523
16 Condensation, evaporation, deposition, and sublimation 525
16.1 Fluxes to and from a single drop 525
16.2 Corrections to growth parameters 528
16.3 Fluxes to a particle with multiple components 540
16.4 Fluxes to a population of particles 540
16.5 Solutions to growth equations 542
16.6 Solving homogeneous nucleation with
condensation 545
16.7 Effects of condensation on coagulation 547
16.8 Ice crystal growth 548
16.9 Summary 550
16.10 Problems 550
x
Contents
17 Chemical equilibrium and dissolution processes 553
17.1 Denitions 553
17.2 Equilibrium reactions 554
17.3 Equilibrium relation and coefcients 558
17.4 Forms of equilibrium-coefcient equations 562
17.5 Mean binary solute activity coefcients 565
17.6 Temperature dependence of binary solute activity
coefcients 567
17.7 Mean mixed solute activity coefcients 568
17.8 The water equation 570
17.9 Solid formation and deliquescence relative
humidity 574
17.10 Example equilibrium problem 575
17.11 Mass-ux iteration method 577
17.12 Analytical equilibrium iteration method 579
17.13 Equilibrium solver results 582
17.14 Nonequilibrium between gases and particles 583
17.15 Summary 594
17.16 Problems 596
18 Cloud thermodynamics and dynamics 598
18.1 Fog and cloud types and formation mechanisms 598
18.2 Moist adiabatic and pseudoadiabatic processes 602
18.3 Cloud development by free convection 606
18.4 Entrainment 608
18.5 Vertical momentum equation in a cloud 610
18.6 Convective available potential energy 612
18.7 Cumulus parameterizations 612
18.8 Cloud microphysics 614
18.9 Summary 642
18.10 Problems 643
19 Irreversible aqueous chemistry 645
19.1 Signicance of aqueous chemical reactions 645
19.2 Mechanisms of converting S(IV) to S(VI) 646
19.3 Diffusion within a drop 652
19.4 Solving growth and aqueous chemical ODEs 654
19.5 Summary 659
19.6 Problems 659
20 Sedimentation, dry deposition, and airsea exchange 661
20.1 Sedimentation 661
20.2 Dry deposition 665
xi
Contents
20.3 Dry deposition and sedimentation calculations 670
20.4 Airsea ux of carbon dioxide and other gases 672
20.5 Summary 679
20.6 Problems 679
21 Model design, application, and testing 681
21.1 Steps in model formulation 681
21.2 Example model simulations 700
21.3 Summary 707
21.4 Problems 707
Appendix A Conversions and constants 709
A.1 Distance conversions 709
A.2 Volume conversions 709
A.3 Mass conversions 709
A.4 Temperature conversions 710
A.5 Force conversions 710
A.6 Pressure conversions 710
A.7 Energy conversions 710
A.8 Power conversions 710
A.9 Speed conversions 710
A.10 Constants 711
Appendix B Tables 714
B.1 Standard atmospheric variables versus altitude 714
B.2 Solar irradiance at the top of the atmosphere 715
B.3 Chemical symbols and structures of gases 716
B.4 Gas-phase reactions 728
B.5 Chemicals involved in equilibrium and aqueous
reactions 738
B.6 Thermodynamic data 740
B.7 Equilibrium reactions and rate coefcients 741
B.8 Irreversible aqueous reactions 743
B.9 Solute activity coefcient data 746
B.10 Water activity data 748
B.11 Surface resistance data 749
B.12 More surface resistance data 751
References 752
Index 784
xii
Preface
Modern atmospheric science is a eld that combines meteorology, physics, mathe-
matics, chemistry, computer sciences, and to a lesser extent geology, biology, micro-
biology, and oceanographic sciences. Until the late 1940s scientic studies of the
atmosphere were limited primarily to studies of the weather. At that time, height-
ened concern about air pollution caused an increase in studies of atmospheric chem-
istry. With the invention of the computer, modeling of weather and air pollution
commenced. Since the late 1940s, the number of meteorological and air-pollution
studies has increased rapidly, and many meteorological and air-pollution models
have merged.
The purposes of this book are to provide (1) a physical understanding of
dynamical meteorology, land- and water-surface processes, radiation, gas chem-
istry, aerosol microphysics and chemistry, and cloud processes, (2) a description of
numerical methods and computational techniques used to simulate these processes,
and (3) a catalog of steps required to construct, apply, and test a numerical model.
The rst chapter of this book gives an overview of model processes and time
scales. Chapter 2 describes atmospheric structure, composition, and thermodynam-
ics. In Chapters 35, basic equations describing dynamical meteorology are derived.
In Chapter 6, numerical methods of solving partial differential equations are dis-
cussed. A technique of solving dynamical meteorological equations is provided in
Chapter 7. In Chapter 8, boundary-layer and ground processes are described. Chap-
ter 9 introduces radiation. Chapters 1012 focus on photochemistry and numerical
methods of solving chemical equations. Chapters 1317 describe aerosol physical
and chemical processes. Chapter 18 discusses cloud thermodynamics and micro-
physics. Chapter 19 discusses aqueous chemistry in aerosol particles and clouds.
Chapter 20 describes sedimentation and dry deposition. Chapter 21 outlines com-
puter model development, application, and testing.
The book is designed as an upper-level undergraduate, graduate, and research
text. The text assumes students have a basic physical science, mathematical, and
computational background. Both Syst` eme Internationale (SI) and centimeter-gram-
second (CGS) units are used. Dynamical meteorologists often use SI units, and
atmospheric chemists often use CGS units. Thus, both unit systems are retained.
Unit and variable conversions are given in Appendix A.
xiii
Acknowledgments
I would like to thank several colleagues who provided comments, suggestions,
and/or corrections relating to the text. In particular, I am indebted to (in alpha-
betical order) A. April, Akio Arakawa, Mary Barth, Jaime Benitez, Merete Bilde,
Steve Bryson, Bob Chateld, Tu-Fu Chen, Johann Feichter, Frank Freedman, Ann
Fridlind, A. V. Gemintern, J. Haigh, Hiroshi Hayami, Roy Harrison, James Holton,
Daniel Jacob, Liming Li, Jinyou Liang, Jin-Sheng Lin, Bernd Kaercher, Gerard
Ketean, Bennert Machenhauer, Ed Measure, Gary Moore, Elmar Reiter, Doug
Rotman, Roberto San Jose, Hjalti Sigurjonsson, Hanwant Singh, Jing Song, Tae-
Joon Song, Amy Stuart, Azadeh Tabazadeh, Roland von Glasow, Chris Walcek,
Thomas Warner, Debra Weisenstein, Don Wuebbles, and Yang Zhang.
xiv
1
Introduction
1.1 BRIEF HISTORY OF METEOROLOGICAL SCIENCES
T
he history of atmospheric sciences begins with weather forecasting. Forecast-
ing originally grew out of three needs for farmers to produce crops, sailors to
survive at sea, and populations to avoid weather-related disasters such as oods.
Every society has forecast wind, rain, and other weather events. Some forecasts
are embodied in platitudes and lore. Virgil stated, Rain and wind increase after
a thunderclap. The Zuni Indians had a saying, If the rst thunder is from the
east, winter is over. Human experiences with the weather have led to more recent
forecast rhymes, such as, Rainbow in morning, sailors take warning. Rainbow at
night, a sailors delight.
Primitive forecasts have also been made based on animal and insect behavior or
the presence of a human ailment. Bird migration was thought to predict oncoming
winds. This correlation has since proved unreliable. Rheumatism, arthritis, and
gout have been associated with the onset of rain, but such ailments are usually
unrelated to the weather. The presence of locusts has correctly been associated
with rainfall in that locusts y downwind until they reach an area of converging
winds, where rain is likely to occur.
In the 1870s, forecasting based on observations and experience became a pro-
fession. Many felt that early professional forecasting was more of an art than a
science, since it was not based on scientic theory. Although the amount of data
available to forecasters was large and increasing, the data were not always used.
Data were gathered by observers who used instruments that measured winds, pres-
sure, temperature, humidity, and rainfall. Many of these instruments had been
developed over the previous two centuries, although ideas and crude technologies
existed prior to that time.
The Greeks, around 430 BC, may have been the rst to measure winds. Yet,
reliable instruments to measure wind force and direction were not developed until
nearly two millennia later. In 1450, the Italian mathematician Leone Battista Alberti
(140472) developed the rst known anemometer, a swinging-plate anemometer
that consisted of a disc placed perpendicular to the wind. It was used to mea-
sure wind speed based on the angle between the disc in its original position and its
displaced position. In 1667 Robert Hooke developed a similar device, the pressure-
plate anemometer, which consisted of a sheet of metal hanging vertically. Wind-
mills were used as early as AD644 in Persia, but the rst spinning-cup anemometer,
1
Introduction
which applies the principle of the windmill to measure wind speed, was not devel-
oped until the nineteenth century. In 1846, the Irish physicist John Thomas Romney
Robinson invented a spinning-cup anemometer that consisted of four hemispher-
ical cups mounted on a vertical axis. In 1892, William Henry Dines invented the
pressure-tube (Dines) anemometer, which is a device that measures wind speed
from the pressure difference arising from wind blowing in a tube versus that blow-
ing across the tube. The pressure difference is proportional to the square of the
wind speed.
In 1643, Evangelista Torricelli (160847) invented the mercury barometer,
becoming the rst to measure air pressure. He lled a glass tube 1.2 m long with
mercury and inverted it onto a dish. He found that only a portion of the mercury
owed from the tube into the dish, and the resulting space above the mercury in
the tube was devoid of air (a vacuum). Thus, Torricelli was also the rst person to
record a sustained vacuum. He suggested that the change in height of the mercury
in the tube each day was caused by a change in atmospheric pressure. Air pressure
balanced the pressure exerted by the column of mercury in the tube, preventing the
mercury from owing freely from the tube. The aneroid barometer, which repre-
sented an advance over the mercury barometer, was not developed adequately until
1843. The aneroid barometer contains no uid. Instead, it measures pressure by
gauging the expansion and contraction of a tightly sealed metal cell that contains
no air.
A third important invention for meteorologists was the thermometer. Prior to
1600, Galileo Galilei (15641642) devised the thermoscope, which estimated tem-
perature change by measuring the expansion of air. The instrument did not have
a scale and was unreliable. Torricellis mercury barometer, which contained uid,
led to the invention of the liquid-in-glass thermometer in Florence in the mid sev-
enteenth century. In the early eighteenth century, useful thermometer scales were
developed by Gabriel Daniel Fahrenheit of Germany (16861736) and Anders
Celsius of Sweden (17011744).
A fourth important invention was the hygrometer, which measures humidity.
In 1450, the German cardinal, philosopher, and administrator Nicolas of Cusa
(Nicolas Cryfts) (140164) described the rst hygrometer with the following:
If someone should hang a good deal of wool, tied together on one end of a
large pair of scales, and should balance it with stones at the other end in a place
where the air is temperate it would be found that the weight of the wool would
increase when the air became more humid, and decrease when the air tended to
dryness.
(Brownawell 2004). In 1481, Leonardo da Vinci (14521519) drew Cryfts
hygrometer in his Codex Atlanticus, using a sponge instead of wool. The purpose
of the hygrometer, according to da Vinci, was
to know the qualities and thickness of the air, and when it is going to rain.
(White 2000). In 1614, Santorio Santorre developed a hygrometer that measured
vapor by the contraction and elongation of cord or lyre strings. Later hygrometers
2
1.1 Brief history of meteorological sciences
were made of wood, seaweed, paper, hair, nylon, and acetate. The hair hygrometer is
still used today, although another instrument, the psychrometer, is more accurate.
A psychrometer consists of two liquid-in-glass thermometers mounted together,
one with a dry bulb and the other with a bulb covered with a moistened cloth.
Following the inventions above, observations of pressure, temperature, humid-
ity, wind force, wind direction, and rainfall became regular. By the nineteenth cen-
tury, weather-station networks and meteorological tables were common. Observers
gathered data and forecasters used the data to predict the weather, but neither
observers nor forecasters applied signicant theory in their work. Theoreticians
studied physical laws of nature but did not take advantage of the abundance of
data available.
One of the rst weather theoreticians was Aristotle, who wrote Meteorologica
about 340 BC. In that text, Aristotle attempted to explain the cause of winds,
clouds, rain, mist, dew, frost, snow, hail, thunder, lightning, thunderstorms, halos,
rainbows, and mock suns. On the subject of winds, he wrote (Lee 1951),
These, then are the most important different winds and their positions. There are
two reasons for there being more winds from the northerly than from the southerly
regions. First, our inhabited region lies toward the north; second, far more rain and
snow is pushed up into this region because the other lies beneath the sun and its
course. These melt and are absorbed by the Earth and when subsequently heated
by the sun and the Earths own heat cause a greater and more extensive exhalation.
On the subject of thunder, he wrote,
Let us now explain lightning and thunder, and then whirlwinds, rewinds and
thunderbolts: for the cause of all of themmust be assumed to be the same. As we have
said, there are two kinds of exhalation, moist and dry; and their combination (air)
contains both potentially. It condenses into cloud, as we have explained before, and
the condensation of clouds is thicker toward their farther limit. Heat when radiated
disperses into the upper region. But any of the dry exhalation that gets trapped when
the air is in process of cooling is forcibly ejected as the clouds condense and in its
course strikes the surrounding clouds, and the noise caused by the impact is what
we call thunder.
Aristotles monograph established a method of qualitatively explaining meteoro-
logical problems. Since Aristotle was incorrect about nearly all his meteorological
conclusions, Meteorologica was never regarded as a signicant work. Aristotle
made observations, as evidenced by diagrams and descriptions in Meteorologica,
but he did not conduct experiments. Lacking experiments, his conclusions, while
rational, were not scientically based.
Aristotles method of rationalizing observations with little or no experiment
governed meteorological theory through the seventeenth century. In 1637, Ren e
Descartes (15961650) wrote Les M et eores, a series of essays attached to Discours
de la M ethode. In some parts of this work, Descartes improved upon Aristotles
treatise by discussing experiments. In other parts, Descartes merely expanded or
3
Introduction
reformulated many of Aristotles explanations. On the subject of northerly winds,
Descartes wrote (Olscamp 1965),
We also observe that the north winds blowprimarily during the day, that they come
from above to below, and that they are very violent, cold and dry. You can see the
explanation of this by considering that the Earth EBFD [referring to a diagram]
is covered with many clouds and mists near the poles E and F, where it is hardly
heated by the sun at all; and that at B, where the sun is immediately overhead, it
excites a quantity of vapors which are quite agitated by the action of its light and
rise into the air very quickly, until they have risen so high that the resistance of their
weight makes it easier for them to swerve, . . .
Like Aristotle, Descartes was incorrect about many explanations. Despite some
of the weaknesses of his work, Descartes is credited with being one of the rst in
meteorological sciences to form hypotheses and then to conduct experiments.
Between the seventeenth and mid nineteenth centuries, knowledge of basic
physics increased, but mathematics and physics were still not used rigorously to
explain atmospheric behavior. In 1860, William Ferrel published a collection of
papers that were the rst to apply mathematical theory to uid motions on a rotat-
ing Earth. This work was the impetus behind the modern-day eld of dynam-
ical meteorology, which uses physics and mathematics to explain atmospheric
motion.
Between 1860 and the early 1900s weather forecasting and theory advanced
along separate paths. In 1903, Vilhelm Bjerknes of Norway (18621951) promul-
gated the idea that weather forecasting should be based on the laws of physics. This
idea was not new, but Bjerknes advanced it further than others (Nebeker 1995).
Bjerknes thought that weather could be described by seven primary variables
pressure, temperature, air density, air water content, and the three components of
wind velocity. He also realized that many of the equations describing the change
in these variables were physical laws already discovered. Such laws included the
continuity equation for air, Newtons second law of motion, the ideal-gas law, the
hydrostatic equation, and the thermodynamic energy equation.
Bjerknes did not believe that meteorological equations could be solved ana-
lytically. He advocated the use of physical principles to operate on graphical
observations to analyze the weather. For example, from a map of observed wind
barbs, which give horizontal wind speeds and directions, he could draw a map of
streamlines (lines of constant direction) and isolines (lines of constant wind speed),
then use graphical differentiation and graphical algebra to determine vertical wind
speeds, which would be drawn on another map. This technique was called graphical
calculus.
Between 1913 and 1919, Lewis Fry Richardson (18811953) developed a differ-
ent method of analyzing the analytical equations describing the weather (Richard-
son 1922). The method involved dividing a region of interest into rectilinear cells
(grid cells), then writing a nite-difference form of the analytical meteorological
equations for each grid cell and solving the equations by hand over all cells. Whereas
Carle Runge and WilhelmKutta developed a method of nite-differencing ordinary
4
1.2 Brief history of air-pollution science
differential equations in the 1890s, Richardson extended nite-differencing (cen-
tral differencing in his case) to partial differential equations and to multiple grid
cells. Richardson was not satised with his solution technique, though, because
data available to test his method were sparse, and predictions from his method
were not accurate. Nevertheless, his was the rst attempt to predict the weather
numerically in detail.
Until the 1940s, much of Richardsons work was ignored because of the lack of
a means to carry out the large number of calculations required to implement his
method. In 1946, John von Neumann (190357), who was associated with work
to build the worlds rst electronic digital computer ENIAC (Electronic Numer-
ical Integrator and Computer), proposed a project to make weather forecasting
its main application. The project was approved, and the rst computer model
of the atmosphere was planned. Among the workers on von Neumanns project
was Jule Charney, who became director of the project in 1948. Charney made
the rst numerical forecast on ENIAC with a one-dimensional model (Charney
1949, 1951). Since then, numerical models of weather prediction have become
more elaborate, and computers have become faster.
1.2 BRIEF HISTORY OF AIR-POLLUTION SCIENCE
Meteorological science is an old and established eld; air-pollution science has
a shorter history. Natural air pollution has occurred on Earth since the planets
formation. Fires, volcanic eruptions, meteorite impacts, and high winds all cause
natural air pollution. Anthropogenic air-pollution problems have existed on urban
scales for centuries and have resulted from burning of wood, vegetation, coal, oil,
natural gas, waste, and chemicals.
In the nineteenth and early twentieth centuries, most air pollution was due
to chimney and smokestack emission of coal and chemical-factory combustion
products. In 1905, Harold Antoine Des Voeux described the combination of smoke
and fog he observed in cities in Great Britain as smog. Smog fromcoal and chemical
combustion resulted in several air pollution episodes that killed thousands of people
between 1850 and 1960. The worst of these was in December 1952, when smog
resulted in over 4000 deaths in London. Pollution resulting fromcoal and chemical-
factory combustion in the presence of fog is commonly referred to as London-type
smog.
In the early twentieth century, the widespread use of automobiles and the
increase in industrial activity increased the prevalence of another type of air pollu-
tion, called photochemical smog. This pollution was most noticeable and formed
almost daily in Los Angeles, California. It became so serious that an Air Pollution
Control District was formed in Los Angeles in 1947 to combat it. The composition
of photochemical smog was not elucidated until 1951, when Arie Haagen-Smit pro-
duced ozone in a laboratory from oxides of nitrogen and reactive organic gases, in
the presence of sunlight and suggested that these gases were the main constituents
of Los Angeles air pollution. Photochemical smog has since been observed in most
cities of the world.
5
Introduction
Before the twentieth century, air pollution was not treated as a science but as a
regulatory problem (Boubel et al. 1994). In Great Britain, emission from furnaces
and steam engines led to the Public Health Act of 1848. Emission of hydrogen
chloride from soap making led to the Alkali Act of 1863. In both cases, pollution
abatement was controlled by agencies. In the nineteenth century, pollution abate-
ment in the United States was delegated to municipalities. In most cases, regula-
tion did not reduce pollution much, but in some cases it led to pollution control
technologies, such as the electrostatic precipitator for reducing particle emission
from smokestacks. In one case, the development of a pollutant-control technology,
the scrubber for removing hydrochloric acid gas from chemical factory emission,
provided incentive for the swift passage of a regulation, the Alkali Act of 1863.
Inventions unrelated to air-pollution regulation reduced some pollution problems.
For example, in the early twentieth century, the advent of the electric motor central-
ized sources of combustion at electric utilities, reducing local air pollution caused
by the steam engine.
1.3 THE MERGING OF AIR-POLLUTION AND
METEOROLOGICAL SCIENCES
In the 1950s, laboratory work was undertaken to understand better the formation
of photochemical and London-type smog. Since the computer was already avail-
able, box models simulating atmospheric chemical reactions were readily imple-
mented. In the 1960s and 1970s, air-pollution models, termed air-quality models,
were expanded to two and three dimensions. Such models included treatment of
emission, transport, gas chemistry, and gas deposition to the ground. Most models
used interpolated elds of meteorological data as inputs. Today, many air quality
models use meteorological elds calculated in real time as inputs.
In the 1970s, atmospheric pollution problems, aside from urban air pollution,
were increasingly recognized. Such problems included regional acid deposition,
global ozone reduction, Antarctic ozone depletion, and global climate change.
Initially, ozone reduction and climate change problems were treated separately
by dynamical meteorologists and atmospheric chemists. More recently, computer
models that incorporate atmospheric chemistry and dynamical meteorology have
been used to study these problems.
1.4 WEATHER, CLIMATE, AND AIR POLLUTION
A model is a mathematical representation of a process. An atmospheric computer
model is a computer-coded representation of dynamical, physical, chemical, and
radiative processes in the atmosphere. In atmospheric models, time-dependent pro-
cesses are mathematically described by ordinary differential equations. Space- and
time-dependent processes are described by partial differential equations. Ordi-
nary and partial differential equations are replaced with nite-difference or other
approximations, then computerized and solved.
6
1.4 Weather, climate, and air pollution
Computer models also solve parameterized and empirical equations. A param-
eterized equation is an equation in which one parameter is expressed in terms of
at least two other parameters. The equation of state, which relates pressure to
temperature and air density, is a parameterized equation. An empirical equation
is an equation in which one parameter is expressed as an empirical function (e.g.,
a polynomial t) of at least one other parameter. Whereas parameterized equa-
tions are derived from insight, empirical equations do not always make physical
sense. Instead, they reproduce observed results under a variety of conditions. In this
text, computer modeling of the atmosphere is discussed. Such modeling requires
solutions to ordinary differential equations, partial differential equations, param-
eterized equations, and empirical equations.
Since the advent of atmospheric computer modeling in 1948, models have been
applied to study weather, climate, and air pollution on urban, regional, and global
scales. Weather is the state of the atmosphere at a given time and place, and climate
is the average of weather events over a long period. Some basic weather variables
include wind speed, wind direction, pressure, temperature, relative humidity, and
rainfall. Standard climate variables include mean annual temperatures and mean
monthly rainfall at a given location or averaged over a region.
Air pollutants are gases, liquids, or solids suspended in the air in high enough
concentration to affect human, animal, or vegetation health, or to erode structures.
Standard air pollution problems include urban smog, acid deposition, Antarctic
ozone depletion, global ozone reduction, and global climate change. Urban smog
is characterized by the local concentration buildup of gases and particles emitted
from automobiles, smokestacks, and other human-made sources. Acid deposition
occurs following long-range transport of sulfur dioxide gas emitted from coal-
red power plants, conversion of the sulfur dioxide to liquid-phase sulfuric acid,
and deposition of sulfuric-acid-related species to the ground by rain or another
means. Acid deposition also occurs when nitric acid gas, produced chemically from
automobile pollutants, dissolves into fog drops, which deposit to the ground or
lungs. This form of acid deposition is acid fog. Acids harm soils, lakes, and forests
and damage structures.
Antarctic ozone depletion and global ozone reduction are caused, to a large
extent, by chlorine and bromine compounds that are emitted anthropogenically
into the atmosphere and break down only after they have diffused to the upper
atmosphere. Ozone reduction increases the intensity of ultraviolet radiation from
the Sun reaching the ground. Some ultraviolet wavelengths destroy microorgan-
isms on the surface of the Earth and cause skin cancer in humans. Global climate
change is characterized by changes in global temperature and rainfall patterns
due to increases in atmospheric carbon dioxide, methane, nitrous oxide, water
vapor, and other gases that absorb infrared radiation. The addition of particles
to the atmosphere may in some cases warm, and in other cases cool, climate as
well.
Historically, meteorological models have been used to simulate weather, climate,
and climate change. Photochemical models have been used to study urban, regional,
and global air-pollution emission, chemistry, aerosol processes, and transport of
7
Introduction
Table 1.1 Scales of atmospheric motion
Scale name Scale dimension Examples
Molecular scale 2 mm
Molecular diffusion, molecular viscosity
Microscale 2 mm2 km Eddies, small plumes, car exhaust, cumulus clouds
Mesoscale 22000 km Gravity waves, thunderstorms, tornados, cloud
clusters, local winds, urban air pollution
Synoptic scale 50010 000 km High- and low-pressure systems, weather fronts,
tropical storms, hurricanes, Antarctic ozone hole
Planetary scale >10 000 km Global wind systems, Rossby (planetary) waves,
stratospheric ozone reduction, global warming
pollutants. Only recently have meteorological models merged with photochemical
models to tackle these problems together.
One purpose of developing a model is to understand better the physical, chemi-
cal, dynamical, and radiative properties of air pollution and meteorology. Asecond
purpose is to improve the model so that it may be used for forecasting. A third
purpose is to develop a tool that can be used for policy making. With an accurate
model, policy makers can try to mitigate pollution problems.
1.5 SCALES OF MOTION
Atmospheric problems can be simulated over a variety of spatial scales. Molecular-
scale motions occur over distances much smaller than 2 mm. Molecular diffu-
sion is an example of a molecular-scale motion. Microscale motions occur over
distances of 2 mm to 2 km. Eddies, or swirling motions of air, are microscale
events. Mesoscale motions, such as thunderstorms, occur over distances of 2
2000 km. The synoptic scale covers motions or events on a scale of 50010 000 km.
High- and low-pressure systems and the Antarctic ozone hole occur over the
synoptic scale. Planetary-scale events are those larger than synoptic-scale events.
Global wind systems are planetary-scale motions. Some phenomena occur on more
than one scale. Acid deposition is a mesoscale and synoptic-scale phenomenon.
Table 1.1 summarizes atmospheric scales and motions or phenomena occurring on
each scale.
1.6 ATMOSPHERIC PROCESSES
Atmospheric models simulate many processes and feedbacks among them.
Figure 1.1 shows a diagram of an air pollutionweatherclimate model that simu-
lates gas, aerosol, cloud, radiative, dynamical, transport, and surface processes.
A gas is an individual atom or molecule suspended in the air in its own phase
state. Gas molecules have diameters on the order of 25 10
10
m.
An aerosol is an ensemble of solid, liquid, or mixed-phase particles suspended
in air. Each particle consists of an aggregate of atoms and/or molecules bonded
together. An aerosol particle is a single particle within an aerosol. Aerosol particles
8
1.6 Atmospheric processes
Emission
Photochemistry
Heterogeneous chemistry
Aerosol nucleation
Condensation/evaporation
Dissolution/evaporation
Dry deposition
Washout
Solar and infrared radiation
Gas, aerosol, cloud absorption
Gas, aerosol, cloud scattering
Heating rates
Actinic fluxes
Visibility
Albedo
Emission
Gas, aerosol, cloud transport in air
Gas, aerosol transport in clouds
Dry deposition/sedimentation
Rainout/washout
Air temperature
Air density
Air pressure
Wind speed and direction
Turbulence
Water vapor
Gas processes
Radiative processes
Aerosol processes
Meteorological processes
Transport processes
Emission
Nucleation
Aerosolaerosol coagulation
Aerosolhydrometeor coagulation
Condensation/evaporation
Dissolution/evaporation
Equilibrium chemistry
Aqueous chemistry
Heterogeneous chemistry
Dry deposition/sedimentation
Rainout/washout
Cloud processes
Condensation/ice deposition
Homogeneous, contact freezing
Melting/evaporation/sublimation
Hydrometeorhydrometeor coag.
Aerosolhydrometeor coagulation
Gas dissolution/aqueous chemistry
Precipitation, rainout, washout
Lightning
Surface processes
Soil, water, sea ice, snow, road,
roof, vegetation temperatures
Surface energy, moisture fluxes
Ocean dynamics
Figure 1.1 Diagram of processes simulated in an air pollutionweather
climate model.
have diameters that range in size from a few tens of gas molecules to 10 mm and
can contain many components, including liquid water.
A hydrometeor is an ensemble of liquid, solid, or mixed-phase predominantly
water-containing particles suspended in or falling through the air. A hydrometeor
particle is a single particle within a hydrometeor. Examples of hydrometeor parti-
cles are cloud drops, ice crystals, raindrops, snowakes, and hailstones. The main
difference between an aerosol particle and a hydrometeor particle is that the latter
contains much more water than the former. Hydrometeor particles generally range
in diameter from 10 m to 10 mm. In this text, the term particle used alone may
refer to an aerosol particle or a hydrometeor particle.
Figure 1.1 lists the major processes affecting gases in the atmosphere. Emission
is the addition of a pollutant to the atmosphere. Photochemistry encompasses gas
kinetic chemistry and photolysis. Gas kinetic chemistry is the process by which
reactant gases collide with each other and transform to product gases. Photoly-
sis is the process by which reactant gases are broken down by sunlight to form
products.
9
Introduction
Gases are also affected by gas-to-particle conversion. Conversion processes
include heterogeneous chemistry, nucleation, condensation/evaporation, dissolu-
tion/evaporation, and deposition/sublimation. Gases react chemically on the sur-
faces of particles to form gas, liquid, or solid products during heterogeneous chem-
istry. Nucleation occurs when gas molecules aggregate and change phase to a liquid
or solid to form a new small aerosol particle or a cluster on an existing particle
surface. Condensation occurs when a gas diffuses to and sticks to the surface of a
particle and changes state to a liquid. Evaporation occurs when a liquid molecule
on a particle surface changes state to a gas and diffuses away from the surface.
Dissolution occurs when a gas molecule diffuses to and dissolves into liquid on the
surface of a particle. Evaporation, in this case, is the opposite of dissolution.
Gases are physically removed from the atmosphere by dry deposition and
washout. Gas dry deposition (different from solid deposition) is a removal process
that occurs when a gas (or particle) impinges upon and sticks to a surface, such
as the ground or a house. Gas washout is the dissolution of a gas in precipitation
(rainfall) that falls to the ground.
Figure 1.1 lists major aerosol processes. Some, such as emission, nucleation,
condensation/evaporation, dissolution/evaporation, and heterogeneous chemistry,
affect gases as well. Aerosolaerosol coagulation occurs when two aerosol parti-
cles collide and coalesce (stick together) to form a third, larger particle. Aerosol
hydrometeor coagulation occurs when an aerosol particle collides and coalesces
with a hydrometeor particle. Equilibriumchemistry is reversible chemistry between
or among liquids, ions, and/or solids within aerosols and hydrometeors. Aqueous
chemistry is irreversible chemistry important in water-containing aerosols and in
hydrometeors.
Aerosol particles are physically removed by dry deposition, sedimentation, rain-
out, and washout. Sedimentation is the process by which particles fall from one
altitude to another or to the surface due to their weight. This differs from aerosol
dry deposition, which occurs when particles contact a surface and stick to the
surface. Aerosol rainout is the growth of cloud drops on aerosol particles and
the eventual removal of the aerosol particle to the surface by precipitation. This
differs from aerosol washout, which is the coagulation of aerosol particles with
precipitation that subsequently falls to the ground.
Clouds are affected by several of the same processes affected by aerosol pro-
cesses. In addition, clouds are affected by ice deposition, which is the growth
of water vapor onto aerosol particles to form ice crystals. Sublimation is the
conversion of the ice crystals back to vapor. During freezing, liquid water within
a hydrometeor changes state to ice. Melting is the reverse. Hydrometeor particles
may coagulate with themselves or with aerosol particles. Lightning occurs when
ice crystals collide with then bounce off of other ice crystals, creating a charge
separation that eventually leads to a ash of light.
Gases, aerosol particles, and hydrometeor particles scatter (redirect) and absorb
solar radiation (emitted by the Sun) and infrared radiation (emitted by the Earth,
atmosphere, and Sun), affecting heating rates (the rates at which the atmosphere
heats or cools), actinic uxes (a parameter used to calculate photolysis rate
10
1.6 Atmospheric processes
coefcients), and visibility (the distance a person can see). Albedo is the reec-
tivity of a surface. It affects radiation in the atmosphere and, in the case of sea ice,
snow, and clouds, is itself affected by pollutants.
Important meteorological variables determined in a model include air temper-
ature, air density, air pressure, wind speed and direction, turbulence, and water
vapor. Winds and turbulence affect transport of gases, aerosol particles, and
hydrometeor particles. Likewise, gases, aerosol particles, and hydrometeor par-
ticles feed back to temperature, pressure, and winds in several ways.
Finally, meteorology and air pollution are affected by surface temperature and
energy/moisture uxes. Surface temperature itself is affected by surface type, some
of which include vegetated and bare soil, water, sea ice, roads, rooftops, and snow.
In a model, meteorological variables are simulated by solving a set of partial dif-
ferential equations and parameterized equations, including the momentum equa-
tion, the thermodynamic energy equation, the continuity equation for air, the equa-
tion of state, and the continuity equation for total water. Heating rates and actinic
uxes are calculated with the radiative-transfer equation. Changes in gas, aerosol
particle, and hydrometeor particle concentration are found by solving the species
continuity equation, which describes transport with partial differential equations
and chemistry/physics with ordinary differential equations.
In sum, weather, climate, and air pollution can be modeled by taking into account
a fairly reasonably well-dened set of physical, chemical, and/or dynamical equa-
tions. In the rest of this book, the processes shown in Fig. 1.1 are examined.
11
2
Atmospheric structure, composition,
and thermodynamics
T
he atmosphere contains a few highly concentrated gases, such as nitrogen,
oxygen, and argon, and many trace gases, among them water vapor, carbon
dioxide, methane, and ozone. All such gases are constituents of air. Important char-
acteristics of air are its pressure, density, and temperature. These parameters vary
with altitude, latitude, longitude, and season and are related to each other by the
equation of state. Two other fundamental equations applicable to the atmosphere
are the ClausiusClapeyron equation and the rst law of thermodynamics. The
ClausiusClapeyron equation relates temperature to the quantity of water vapor
over a surface at saturation. The rst law of thermodynamics relates the temper-
ature change of a gas to energy transfer and the change in work. In this chap-
ter, atmospheric variables and gases are discussed, and basic equations describing
atmospheric physics and thermodynamics are introduced.
2.1 PRESSURE, DENSITY, AND COMPOSITION
In the Earths atmosphere, air density, pressure, and temperature change with alti-
tude. Air density is the mass of air per unit volume of air, where the mass of air is
summed over all gases, aerosol particles, and hydrometeor particles. Since the mass
concentration of gas in the air is 500 times that of hydrometeor particles in the
thickest cloud and ten million times that of aerosol particles in polluted air (Table
13.1), air density is accurately determined from the mass of gas only. Air density
decreases exponentially with altitude, as shown in Fig. 2.1(a). Air density is the
greatest near the surface since atmospheric mass is concentrated near the surface.
Air pressure is ideally the weight (force) of air above a horizontal plane, divided
by the area of the plane. This type of air pressure is called hydrostatic air pres-
sure, which is the pressure due solely to the weight of air in a column above a
given altitude. The term hydrostatic means uid at rest. Air pressure is hydro-
static only if air is not accelerating vertically, which occurs either if it is at rest
or if it has a constant vertical speed. The assumption that air pressure is hydro-
static is reasonable when air pressure is averaged over a large horizontal area
(>23 km in diameter) and outside of a storm system, since vertical accelerations
in such cases are generally small. Over small areas (-23 km in diameter) and in
individual clouds, vertical accelerations can be large. When air accelerates verti-
cally, air pressure is nonhydrostatic. Nonhydrostatic air pressure is discussed in
Section 5.1.3.
12
2.1 Pressure, density, and composition
0
20
40
60
80
100
0 0.4 0.8 1.2
A
l
t
i
t
u
d
e

a
b
o
v
e

s
e
a

l
e
v
e
l

(
k
m
)
Air density (kg m
3
)
(a)
0
20
40
60
80
100
0 200 400 600 800 1000
A
l
t
i
t
u
d
e

a
b
o
v
e

s
e
a

l
e
v
e
l

(
k
m
)
Air pressure (hPa)
1 hPa (above 99.9%)
10 hPa (above 99%)
100 hPa (above 90%)
500 hPa (above 50%)
(b)
0
20
40
60
80
100
9.5 9.6 9.7 9.8 9.9
A
l
t
i
t
u
d
e

a
b
o
v
e

s
e
a

l
e
v
e
l

(
k
m
)
Gravity (m s
2
)
(c)
Figure 2.1 Variation of (a) air density, (b) air pressure, and (c) gravita-
tional acceleration versus altitude in the Earths lower atmosphere. The
pressure diagram shows that 99.9 percent of the atmosphere lies below
an altitude of about 48 km (1 hPa), and 50 percent lies below about
5.5 km (500 hPa).
Hydrostatic air pressure at any altitude z (m) above sea level is found by inte-
grating the product of air density and gravity from the top of the atmosphere to
altitude z with
p
a
(z) =
a
(z)g(z)dz (2.1)
where p
a
(z) is air pressure as a function of altitude (pascal, where 1 Pa =1 Nm
2
=
1 kg m
1
s
2
= 0.01 hPa = 0.01 millibar or mb),
a
(z) is air density as a function
of altitude (kg m
3
) from Fig. 2.1(a), and g(z) is gravitational acceleration (ms
2
).
Alternative units for air pressure are given in Appendix Table A.6. Figures 2.1(b)
and (c) show variations in pressure and gravitational acceleration, respectively,
with altitude in the atmosphere. Tabulated values of these variables are given in
Appendix Table B.1.
13
Atmospheric structure, composition, and thermodynamics
Table 2.1 Average composition of the lowest 100 km in the
Earths atmosphere
Volume mixing ratio
Gas (percent) (ppmv)
Fixed gases
Nitrogen (N
2
) 78.08 780 000
Oxygen (O
2
) 20.95 209 500
Argon (Ar) 0.93 9 300
Neon (Ne) 0.0015 15
Helium (He) 0.0005 5
Krypton (Kr) 0.0001 1
Xenon (Xe) 0.000005 0.05
Variable gases
Water vapor (H
2
O) 0.000014.0 0.140 000
Carbon dioxide (CO
2
) 0.0375 375
Methane (CH
4
) 0.00017 1.8
Ozone (O
3
) 0.0000030.001 0.0310
Figure 2.1(b) shows that, as altitude (z) increases, air pressure decreases expo-
nentially. Air pressure decreases with increasing altitude because less air lies above
a higher altitude than a lower altitude. Air pressure decreases exponentially with
increasing altitude because density, used to derive pressure, decreases exponentially
with increasing altitude (Fig. 2.1(a)).
Figure 2.1(b) also shows that 50 percent of atmospheric mass lies between sea
level and 5.5 km. About 99.9 percent of mass lies below about 48 km. The Earths
radius is approximately 6370 km. Thus, almost all of Earths atmosphere lies in a
layer thinner than 1 percent of the radius of the Earth.
Standard sea-level surface pressure is 1013 hPa (or 1013 mb or 760 mmHg). The
sea-level pressure at a given location and time typically varies by +10 to 20 hPa
fromstandard sea-level pressure. In a strong low-pressure system, such as at the cen-
ter of a hurricane, the actual sea-level pressure may be more than 50 hPa lower than
standard sea-level pressure. At the surface of the Earth, which averages 231.4 m
above sea level globally, gravitational acceleration is 9.8060 ms
2
. Gravitational
acceleration is discussed further in Section 4.2.3.
2.1.1 Fixed gases
Table 2.1 gives the basic composition of the bottom 100 km of the Earths atmo-
sphere, called the homosphere. In this region, the primary gases are molecular
nitrogen (N
2
) and molecular oxygen (O
2
), which together make up over 99 per-
cent of all air molecules. Argon (Ar), a chemically inert gas, makes up most of the
remaining 1 percent. Nitrogen, oxygen, and argon are xed gases in that their vol-
ume mixing ratios (number of molecules of each gas divided by the total number
of molecules of dry air) do not change substantially in time or space. Fixed gases
are well mixed in the homosphere. At any given altitude, oxygen makes up about
14
20.95 percent and nitrogen makes up about 78.08 percent of all non-water gas
molecules by volume (23.17 and 75.55 percent, respectively, by mass).
2.1.1.1 Molecular nitrogen
Gas-phase molecular nitrogen (N
2
) is produced biologically in soils by denitrica-
tion, a two-step process carried out by denitrifying bacteria in anaerobic (oxygen-
depleted) soils. In the rst step, the bacteria reduce the nitrate ion (NO
3
) to the
nitrite ion (NO
2
). In the second, they reduce the nitrite ion to molecular nitrogen

and some nitrous oxide gas (N
2
O).
Molecular nitrogen is removed from the air by nitrogen xation. This process
is carried out by bacteria, such as Rhizobium, Azotobacter, and Beijerinckia, in
aerobic (oxygen-rich) environments, and produces the ammonium ion (NH
4
+
).
Ammonium in soils is also produced by the bacterial decomposition of organic
matter during ammonication. Fertilizer application also adds ammonium to soils.
The ammoniumion regenerates the nitrate ion in soils during nitrication, which
is a two-step aerobic process. In the rst step, nitrosofying (nitrite-forming) bacteria
produce nitrite from ammonium. In the second step, nitrifying (nitrate-forming)
bacteria produce nitrate from nitrite.
Because N
2
does not react signicantly in the atmosphere and because its removal
by nitrogen xation is slower than its production by denitrication, its atmospheric
concentration has built up over time.
2.1.1.2 Molecular oxygen
Gas-phase molecular oxygen is produced by green-plant photosynthesis, which
occurs when carbon dioxide gas (CO
2
) reacts with water (H
2
O) in the presence
of sunlight and green-pigmented chlorophylls. Chlorophylls exist in plants, trees,
blue-green algae, and certain bacteria. They appear green because they absorb red
and blue wavelengths of visible light and reect green wavelengths. Products of
photosynthesis reactions are carbohydrates and molecular oxygen gas. One pho-
tosynthetic reaction is
6CO
2
+6H
2
O+h C
6
H
12
O
6
+6O
2
(2.2)
where h is a photon of radiation, and C
6
H
12
O
6
is glucose. Some bacteria that live
in anaerobic environments photosynthesize carbon dioxide with hydrogen sulde
(H
2
S) instead of with water to produce organic material and elemental sulfur (S).
This type of photosynthesis is anoxygenic photosynthesis and predates the onset
of green-plant photosynthesis.
2.1.1.3 Argon and others
Argon is a noble gas that is colorless and odorless. Like other noble gases, argon
is inert and does not react chemically. It forms from the radioactive decay of
15
potassium. Other xed but inert gases present in trace concentrations are neon (Ne),
helium (He), krypton (Kr), and xenon (Xe). The source of helium, krypton, and
xenon is radioactive decay of elements in the Earths crust, and the source of neon
is volcanic outgassing. The mixing ratios of these gases are given in Table 2.1.
2.1.2 Variable gases
Gases in the atmosphere whose volume mixing ratios change signicantly in time
and space are variable gases. Thousands of variable gases are present in the atmo-
sphere. Several dozen are very important. Four of these, water vapor, carbon diox-
ide, methane (CH
4
), and ozone (O
3
), are discussed below.
2.1.2.1 Water vapor
Major sources of water vapor to the air are evaporation from soils, lakes, streams,
rivers, and oceans, sublimation from glaciers, sea ice, and snow packs, and tran-
spiration from plant leaves. Water vapor is also produced during fuel combustion
and many gas-phase chemical reactions. Approximately 85 percent of water in
the atmosphere originates from ocean surface evaporation. Sinks of water vapor
are condensation to the liquid phase, ice deposition to the solid phase, transfer to
the ocean and other surfaces, and gas-phase chemical reaction. The mixing ratio
of water vapor varies with location. When temperatures are low, water vapor
readily condenses as a liquid or deposits as ice, so its gas-phase mixing ratio
is low. Over the North and South Poles, water vapor mixing ratios are almost
zero. When temperatures are high, liquid water readily evaporates and ice readily
sublimates to the gas phase, so water vapor mixing ratios are high. Over equa-
torial waters, where temperatures are high and evaporation from ocean surfaces
occurs readily, the atmosphere contains up to 4 percent or more water vapor by
volume. Water vapor is not only a greenhouse gas (a gas that readily absorbs
thermal-infrared radiation), but also a chemical reactant and carrier of latent
heat.
2.1.2.2 Carbon dioxide
Carbon dioxide gas is produced by cellular respiration in plants and trees, biolog-
ical decomposition of dead organic matter, evaporation from the oceans, volcanic
outgassing, and fossil-fuel combustion. Cellular respiration occurs when oxygen
reacts with carbohydrates in the presence of enzymes in living cells to produce CO
2
,
H
2
O, and energy. The reverse of (2.2) is a cellular respiration reaction. Biological
decomposition occurs when bacteria and other organisms convert dead organic
matter to CO
2
and H
2
O.
Like water vapor, CO
2
is a greenhouse gas. Unlike water vapor, CO
2
does
not react chemically in the atmosphere. Its lifetime against chemical destruction
is approximately 100200 years. CO
2
is removed more readily by green-plant
16
310
320
330
340
350
360
370
380
1960 1970 1980 1990 2000
C
O
2

m
i
x
i
n
g

r
a
t
i
o

(
p
p
m
v
)
Year
Figure 2.2 Yearly and seasonal uctuations in CO
2
mixing ratio since 1958. Data from C. D. Keeling at
the Mauna Loa Observatory, Hawaii.
photosynthesis and dissolution into ocean water and raindrops. These processes
reduce its overall lifetime in the air to 3095 years.
Average global CO
2
mixing ratios have increased fromapproximately 275 parts
per million by volume (ppmv) in the mid 1700s to approximately 375 ppmv in
2004. Figure 2.2 shows how observed CO
2
mixing ratios have increased steadily
since 1958. The yearly increases are due to increased rates of CO
2
emission from
fossil-fuel combustion. The seasonal uctuation in CO
2
mixing ratio is due to
photosynthesis and biological decomposition. When annual plants grow in the
spring and summer, they remove CO
2
from air by photosynthesis. When they die
in the fall and winter, they return CO
2
by biological decomposition.
2.1.2.3 Methane
Methane, the main component of natural gas, is a variable gas. It is produced in
anaerobic environments, where methane-producing bacteria (methanogens) con-
sume organic material and excrete methane. Ripe anaerobic environments include
rice paddies, landlls, wetlands, and the digestive tracts of cattle, sheep, and ter-
mites. Methane is also produced in the ground fromthe decomposition of fossilized
carbon. This methane often leaks to the atmosphere or is harnessed and used as a
source of energy. Additional methane is produced from biomass burning and as a
by-product of atmospheric chemical reaction. Sinks of methane include slowchem-
ical reaction and dry deposition. Methanes mixing ratio in the lower atmosphere
is near 1.8 ppmv, which is an increase from about 0.7 ppmv in the mid 1700s.
Its mixing ratio has increased steadily due to increased farming and landll usage.
Methane is important because it is a greenhouse gas that absorbs infrared radiation
emitted by the Earth 21 times more efciently, molecule for molecule, than does
carbon dioxide. Mixing ratios of carbon dioxide, however, are much larger than
are those of methane.
17
2.1.2.4 Ozone
Ozone is a trace gas formed by photochemical reaction and is not emitted into the
atmosphere. In the stratosphere, it is produced following photolysis of molecular
oxygen. Near the surface of Earth, it is produced following photolysis of nitrogen
dioxide (NO
2
). Photolysis of molecular oxygen does not occur near the surface,
since the wavelengths of radiation required to break apart oxygen are absorbed
by molecular oxygen and ozone in the upper atmosphere. In urban regions, ozone
production is enhanced by organic gases. Typical ozone mixing ratios in urban air,
rural surface air, and stratospheric air are 0.1, 0.04, and 10 ppmv, respectively.
Thus, ozones mixing ratios are the highest in the stratosphere, the location of
which is shown in Fig. 2.4 (Section 2.2.2.1).
2.2 TEMPERATURE STRUCTURE
Temperature is a measure of the kinetic energy of an air molecule. At a given
temperature, the speeds at which air molecules travel are distributed statistically
about a Maxwellian distribution. Fromthis distribution, one can dene an average
thermal speed ( v
a
), a root-mean-square speed (v
rms
), and a most probable speed
(v
p
) of an air molecule (m s
1
). These speeds are related to absolute temperature
(T, K) by
v
a
=
8k
B
T

M
v
rms
=
3k
B
T
M
v
p
=
2k
B
T
M
(2.3)
respectively, where k
B
is Boltzmanns constant (1.380658 10
23
kg m
2
s
2
K
1
molec.
1
; alternative units are given in Appendix A) and

M is the average mass of
one air molecule (4.8096 10
26
kg molec.
1
). Equation (2.3) can be rewritten
as
4
k
B
T =
1
2
M v
2
a
3
2
k
B
T =
1
2
Mv
2
rms
k
B
T =
1
2
Mv
2
p
(2.4)
The right side of each equation in (2.4) is kinetic energy. When dened in terms of
v
a
, temperature is proportional to the kinetic energy of an air molecule traveling at
its average speed. When dened in terms of v
rms
, temperature is proportional to the
average kinetic energy among all air molecules since v
2
rms
is determined by summing
the squares of all individual speeds then dividing by the total number of speeds.
When dened in terms of v
p
, temperature is proportional to the kinetic energy of
an air molecule traveling at its most probable speed. In this text, temperature is
dened in terms of v
a
.
18
2.2 Temperature structure
Example 2.1
What is the average thermal speed, the root-mean-square speed, and the
most probable speed of an air molecule when T = 300 K? What about at
200 K? Although these speeds are faster than those of an airplane, why do air
molecules hardly move relative to airplanes over the same time increment?
SOLUTION
At 300 K, :
a
= 468.3 ms
1
(1685 kmhr
1
), :
rms
= 508.3 ms
1
, and :
p
=
415.0 ms
1
. At 200 K, :
a
= 382.4 ms
1
(1376 kmhr
1
), :
rms
= 415.0 ms
1
,
and :
p
=338.9 ms
1
. Although air molecules move quickly, they continuously
collide with and are redirected by other air molecules, so their net migration
rate is slow, particularly in the lower atmosphere. Airplanes, on the other hand
are not redirected when they collide with air molecules.
The bottom 100 km of Earths atmosphere, the homosphere, is the lower atmo-
sphere and is divided into four major regions in which temperature changes with
altitude. These are, frombottomto top, the troposphere, stratosphere, mesosphere,
and thermosphere. The troposphere is divided into the boundary layer, which is the
region from the surface to about 5003000-m altitude, and the free troposphere,
which is the rest of the troposphere.
2.2.1 Boundary layer
The boundary layer is the portion of the troposphere inuenced by the Earths
surface and that responds to surface forcing with a time scale of about an hour or
less (Stull 1988). The free troposphere is inuenced by the boundary layer, but on
a longer time scale. Temperature varies signicantly in the boundary layer during
the day and between day and night. Variations are weaker in the free troposphere.
Temperatures in the boundary are affected by the specic heats of soil and air and
by energy transfer processes, such as conduction, radiation, mechanical turbulence,
thermal turbulence, and advection. These factors are described below.
2.2.1.1 Specic heat
Specic heat capacity (specic heat, for short) is the energy required to increase
the temperature of 1 gram of a substance 1 degree Celsius (
C). Soil has a lower

specic heat than has liquid water. During the day, the addition of a xed amount
of sunlight increases soil temperature more than it increases water temperature.
At night, emission of a xed amount of infrared radiation decreases soil tempera-
ture more than it decreases water temperature. Thus, between day and night, soil
temperature varies more than does water temperature.
19
Table 2.2 Specic heats and thermal conductivities of four media
Specic heat Thermal conductivity at
Substance (J kg
1
K
1
) 298 K (J m
1
s
1
K
1
)
Dry air at constant pressure 1004.67 0.0256
Liquid water 4185.5 0.6
Clay 1360 0.920
Dry sand 827 0.298
Specic heat varies not only between land and water, but also among soil types.
Table 2.2 shows that the specic heat of clay is greater than is that of sand, and
the specic heat of liquid water is greater than is that of clay or sand. If all else is
the same, sandy soil heats to a higher temperature than does clay soil, and dry soil
heats to a higher temperature than does wet soil during the day. Dry, sandy soils
cool to a greater extent than do wet, clayey soils at night.
2.2.1.2 Conduction
Specic heat is a property of a material that affects its temperature change. Temper-
ature is also affected by processes that transfer energy within or between materials.
One such process is conduction. Conduction is the passage of energy from one
molecule to the next in a medium (the conductor). The medium, as a whole, expe-
riences no molecular movement. Conduction occurs through soil, air, and particles.
Conduction affects ground temperature by transferring energy between the soil sur-
face and bottom molecular layers of the atmosphere, and between the soil surface
and molecules of soil just below the surface. The rate of a materials conduction is
determined by its thermal conductivity, which quanties the rate of ow of ther-
mal energy through a material in the presence of a temperature gradient. Empirical
expressions for the thermal conductivities of dry air and water vapor (J m
1
s
1
K
1
) are
d
0.023807 +7.1128 10
5
(T 273.15) (2.5)
v
0.015606 +8.3680 10
5
(T 273.15) (2.6)
respectively, which are used in an interpolation equation to give the thermal con-
ductivity of moist air (dry air plus water vapor)
a

d
1.17 1.02
n
v
n
v
+n
d
(2.7)
(Pruppacher and Klett 1997). Here, n
v
and n
d
are the number of moles of water
vapor and dry air, respectively, and T is temperature (K). Under atmospheric con-
ditions, the thermal conductivity of moist air is not much different from that of dry
air. Table 2.2 shows that liquid water, clay, and dry sand are more conductive than
20
is dry air. Clay is more conductive and dry sand is less conductive than is liquid
water.
The ux of energy through the atmosphere due to conduction can be estimated
with the conductive heat ux equation,
H
c
=
a
LT
Lz
(2.8)
(W m
2
), where LT (K) is the change in temperature over an incremental height
Lz (m). Adjacent to the ground, molecules of soil, water, and other surface elements
transfer energy by conduction to molecules of air overlying the surface. Since the
temperature gradient (LTLz) between the surface and a thin (e.g., 1 mm) layer of
air just above the surface is large, the conductive heat ux at the ground is large.
Above the ground, temperature gradients are small, and the conductive heat ux
is much smaller than at the ground.
Example 2.2
Find the conductive heat ux through a 1-mm thin layer of air touching
the surface if T = 298 K, LT = 12 K, and assuming the air is dry. Do a
similar calculation for the free troposphere, where T = 273 K and LTLz =
6.5 K km
1
.
SOLUTION
Near the surface,
a

d
= 0.0256 J m
1
s
1
K
1
, giving a conductive heat ux
of H
c
= 307 W m
2
. In the free troposphere,
a

d
= 0.0238 J m
1
s
1
K
1
,
giving H
c
= 1.5 10
4
Wm
2
. Thus, heat conduction through the air is impor-
tant only at the ground.
2.2.1.3 Radiation
Radiation is the transfer of energy by electromagnetic waves, which do not require
a medium, such as air, for their transmission. Solar radiation is relatively short-
wavelength radiation emitted by the Sun. Thermal-infrared radiation is relatively
long-wavelength radiation emitted by the Earth, atmosphere, and clouds. The
Earths surface receives solar radiation during the day only, but its surface and
atmosphere emit thermal-infrared radiation during day and night. Radiation is an
important energy transfer process and is discussed in Chapter 9.
2.2.1.4 Mechanical turbulence and forced convection
Convection is a predominantly vertical motion that results in the transport and
mixing of atmospheric properties. Forced convection is vertical motion produced
by mechanical means, such as mechanical turbulence. Mechanical turbulence arises
when winds travel over objects protruding from a surface, producing swirling
21
motions of air, or eddies. Turbulence is the effect of groups of eddies of dif-
ferent size. Turbulence mixes energy and other variables vertically and horizon-
tally. Strong winds produce strong eddies and turbulence. Turbulence from wind-
generated eddies is mechanical turbulence.
When mechanical turbulence is the dominant process of vertical motion in the
boundary layer, the boundary layer is in a state of forced convection. Mechanical
turbulence is only one type of forced convection; forced convection also occurs
when air rises along a topographical barrier or weather front, or when horizontal
winds converge and rise.
2.2.1.5 Thermal turbulence and free convection
Free convection is a predominantly vertical motion produced by buoyancy. The
boundary layer is in a state of free convection when thermal turbulence is the dom-
inant process of vertical motion. Thermal turbulence occurs when the Sun heats the
ground differentially, creating thermals. Differential heating occurs because clouds
or hills block the sun in some areas but not in others or because different surfaces
lie at different angles. Thermals are parcels of air that rise buoyantly, generally over
land during the day. When the Sun heats the ground, conduction transfers energy
from the ground to molecules of air adjacent to the ground. The warmed air above
the ground rises buoyantly and expands. Cooler air from nearby is drawn in near
the surface to replace the rising air. The cooler air heats by conduction and rises.
Free convection differs fromconduction in that free convection is the mass move-
ment of air molecules containing energy due to density difference, and conduction
is the transfer of energy from molecule to molecule. Free convection occurs most
readily over land when the sky is cloud-free and winds are light.
2.2.1.6 Advection and other factors
Advection is the horizontal propagation of the mean wind. Horizontal winds advect
energy spatially just as they advect gases and particles. Advection is responsible for
transferring energy on small and large scales. Other processes that affect tempera-
ture in the boundary layer are emissivity, albedo, pressure systems, and length of
day.
Emissivity is the ratio of the radiation emitted by an object to the radiation
emitted by a perfect emitter. Sand has an emissivity of 0.840.91, and clay has an
emissivity of 0.90.98. The higher its emissivity, the faster a surface, such as soil,
cools at night. Albedo (or reectivity) is the ratio of reected radiation to incident
radiation. For dry sand, the albedo varies from 20 to 40 percent. For clay, it varies
from 5 to 20 percent. Thus, sand reects more solar radiation during the day and
emits less infrared radiation at night than does clay, counteracting some of the
effects of the low specic heat and thermal conductivity of sand.
Large-scale pressure systems also affect temperatures in the boundary layer.
Within a large-scale high-pressure system, air descends and warms, often on top of
cooler surface air, creating an inversion, which is an increase in temperature with
22
Cloud
layer
Entrainment zone /
Inversion layer
Free troposphere
Neutral convective
mixed layer
Surface layer
Subcloud layer
Daytime temperature
A
l
t
i
t
u
d
e
B
o
u
n
d
a
r
y

l
a
y
e
r
(a)
Entrainment zone /
Inversion layer
Free troposphere
Surface layer
Nighttime temperature
A
l
t
i
t
u
d
e
Neutral
residual layer
Stable
boundary
layer
B
o
u
n
d
a
r
y

l
a
y
e
r
(b)
Figure 2.3 Variation of temperature with height during the (a) day and (b)
night in the atmospheric boundary layer over land under a high-pressure sys-
tem. Adapted from Stull (1988).
increasing height. An inversion associated with a large-scale pressure system is a
large-scale subsidence inversion. As air warms when it sinks in a subsidence inver-
sion, it evaporates clouds, increasing the sunlight reaching the ground, increasing
ground temperature. Within a large-scale low-pressure system, air rises and cools,
often creating clouds and decreasing sunlight reaching the surface.
Temperature is also affected by length of day. Longer days produce longer peri-
ods of surface heating by sunlight, and longer nights produce longer periods of
cooling due to lack of sunlight and continuous thermal-infrared emission.
2.2.1.7 Boundary-layer characteristics
Figures 2.3(a) and (b) show idealized temperature variations in the boundary layer
over land during the day and night, respectively, under a high-pressure system.
During the day, the boundary layer consists of a surface layer, a convective mixed
layer, and an entrainment zone. The surface layer is a region of strong wind shear
that comprises the bottom 10 percent of the boundary layer. Since the boundary-
layer depth ranges from 500 to 3000 m, the surface layer is about 50300 m
thick.
Over land during the day, temperature decreases rapidly with increasing altitude
in the surface layer but less so in the mixed layer. In the surface layer, the strong
temperature gradient is caused by rapid solar heating of the ground. The temper-
ature gradient is usually so strong that air adjacent to the ground buoyantly rises
and accelerates into the mixed layer. The atmosphere is called unstably stratied
when it exhibits the strong decrease in temperature with increasing height required
for air parcels to rise and accelerate buoyantly.
In the mixed layer, the temperature gradient is not strong enough to allow unre-
strained convection but not weak enough to prevent some convection. Under such
23
a condition, the atmosphere is neutrally stratied, and parcels of air can mix up or
down but do not accelerate in either direction.
When a high-pressure systemis present, a large-scale subsidence inversion resides
above the mixed layer. Environments in which temperatures increase with increas-
ing height (an inversion) or slightly decrease with increasing height are stably strati-
ed. Thermals originating in the surface layer or mixed layer cannot easily penetrate
buoyantly through an inversion. Some mixing (entrainment) between an inversion
layer and mixed layer always occurs; thus, the inversion layer is also called an
entrainment zone.
Other features of the daytime boundary layer are the cloud layer and subcloud
layer. A region in which clouds appear in the boundary layer is the cloud layer,
and the region underneath is the subcloud layer. A cloud forms if rising air in a
thermal cools sufciently. An inversion may prevent a cloud from rising past the
mixed layer.
During the night, the surface cools radiatively, causing temperature to increase
with increasing height fromthe ground, creating a surface inversion. Eventual cool-
ing at the top of the surface layer cools the bottomof the mixed layer, increasing the
stability of the mixed layer at its base. The portion of the daytime mixed layer that
is stable at night is the stable (nocturnal) boundary layer. The remaining portion of
the mixed layer, which stays neutrally stratied, is the residual layer. Because ther-
mals do not form at night, the residual layer does not undergo signicant change,
except at its base. At night, the nocturnal boundary layer grows, eroding the base
of the residual layer. The top of the residual layer is not affected by this growth.
Over the ocean, the boundary layer is inuenced more by large-scale pressure
systems than by thermal or mechanical turbulence. Since water temperature does
not change signicantly during the day, thermal turbulence over the ocean is not
so important as over land. Since the ocean surface is relatively smooth, mechanical
turbulence is also weaker than over land. Large-scale high-pressure systems still
cause subsidence inversions to form over the ocean.
2.2.2 Free atmosphere
2.2.2.1 Troposphere
The free troposphere lies above the boundary layer. Figure 2.4 shows a standard
prole of the temperature structure of the lower atmosphere, ignoring the bound-
ary layer. The troposphere, which is the bottom layer of the lower atmosphere,
is the region extending from the surface in which the temperature, on average,
decreases with increasing altitude. The average rate of temperature decrease in the
free troposphere (above the boundary layer) is about 6.5 K km
1
.
The temperature decreases with increasing altitude in the free troposphere for
the following simplied reason: Earths surface receives energy from the Sun daily,
heating the ground. Simultaneously, the top of the troposphere continuously radi-
ates energy upward, cooling the upper troposphere. The troposphere, itself, has rel-
atively little capacity to absorb solar energy; thus, it relies on radiative, turbulent,
24
0
10
20
30
40
50
60
70
80
90
100
180 200 220 240 260 280 300
1013
265
55
12
2.9
0.8
0.22
0.052
0.011
0.0018
0.00032
A
l
t
i
t
u
d
e

(
k
m
)
Temperature (K)
Tropopause
Stratopause
Mesopause
Stratosphere
Troposphere
Mesosphere
Thermosphere
Ozone
layer
P
r
e
s
s
u
r
e

(
h
P
a
)
Figure 2.4 Temperature structure of the Earths lower atmo-
sphere, ignoring the boundary layer.
and conductive energy transfer from the ground to maintain its temperature. The
most important of these transfer processes is (thermal-infrared) radiation transfer,
followed by turbulence, then conduction. The last two processes transfer energy
to the boundary layer only. Thermal-infrared radiation emitted by the ground is
absorbed by selective gases in the air, including water vapor and carbon dioxide.
These and other gases are most concentrated near the surface, thus they have the
rst opportunity to absorb the radiation. Once they absorb the radiation, they
reemit half of it upward and the other half downward, but their emission is at a
lower temperature than is the emission of the radiation they absorbed. The lower
the temperature of radiant emission, the less intense the emission. Thus, not only
does less thermal-infrared radiation penetrate to higher than to lower altitudes, but
the radiation that does penetrate is less intense than at lower altitudes. Because less
radiation is available for absorption with increasing altitude, temperatures decrease
with increasing altitude in the troposphere.
Figures 2.5(a) and (b) show global latitudealtitude contour plots of zonally
averaged temperatures for a generic January and July, respectively. A zonally aver-
aged temperature is found from a set of temperatures, averaged over all longitudes
at a given latitude and altitude. The gures indicate that near-surface tropospheric
temperatures decrease from the Equator to high latitudes, which is expected, since
the Earth receives the greatest quantity of incident solar radiation near the Equator.
The tropopause is the upper boundary of the troposphere. It is dened by the
World Meteorological Organization (WMO) as the lowest altitude at which the
lapse rate (rate of decrease of temperature with increasing height) decreases to
2 K km
1
or less, and at which the lapse rate, averaged between this altitude and
any altitude within the next 2 km, does not exceed 2 Kkm
1
. Above the tropopause
base, the temperature is relatively constant with increasing altitude (isothermal)
before it increases with increasing height in the stratosphere.
Figures 2.5(a) and (b) indicate that the tropopause height decreases from 15
18 km near the Equator to 8 km near the poles. Because temperature decreases
with increasing height in the troposphere, and because tropopause height is higher
25
240
280
220
210
200
160
220
200
240
260
180
220
210
80 60 40 20 0 20 40 60 80
100
80
60
40
20
0
Latitude (deg)
A
l
t
i
t
u
d
e

(
k
m
)
(a)
140
160
180
200
220
240
220
260
280 260
240
220
200
210
80604020 0 20 40 60 80
100
80
60
40
20
0
Latitude (deg)
A
l
t
i
t
u
d
e

(
k
m
)
(b)
Figure 2.5 Zonally and monthly averaged temperatures for (a) January and
(b) July. Data for the plots were compiled by Fleming et al. (1988).
over the Equator than over the poles, temperature decreases to a lower value over
the Equator than over the poles. As a result, minimum tropopause temperatures
occur over equatorial regions, as shown in Figs. 2.5(a) and (b). Tropopause heights
are the highest over the Equator because strong vertical motions over the Equator
raise the base of the ozone layer and force ozone to spread horizontally to higher
latitudes. Since ozone is responsible for warming above the tropopause, pushing
ozone to a greater height increases the altitude at which warming begins. Near
the poles, downward motion pushes ozone downward. The resulting latitudinal
gradient of the ozone-layer base decreases the tropopause height from the Equator
to the poles.
Tropopause temperature is also affected by water vapor. Water vapor absorbs
the Earths thermal-infrared radiation, preventing some of it from reaching the
tropopause. Water-vapor mixing ratios are much higher over the Equator than
over the poles. The high mixing ratios near the surface over the Equator enhance
cooling of the tropopause above.
2.2.2.2 Stratosphere
The stratosphere is a region of the atmosphere above the troposphere that contains
90 percent of the Earths ozone and in which temperature increases with increasing
height. The temperature prole in the stratosphere is caused by ozone interactions
with radiation, discussed below.
Ozone is produced when molecular oxygen absorbs ultraviolet radiation and
photodissociates by the reactions,
O
2
+h
(
1
D) +
- 175 nm (2.9)
O
2
+h
+

O
175 - - 245 nm (2.10)

where h is a photon of radiation, O(
1
D) is excited atomic oxygen, O [=O(
3
P)] is
ground state atomic oxygen, the dots identify the valence of an atom (Chapter 10),
26
and is a wavelength of radiation affecting the reaction. O(
1
D) produced from
(2.9) rapidly converts to O via
O(
1
D) +M
O+M (2.11)
where Mprovides collisional energy for the reaction, but is not created or destroyed
by it. Because N
2
and O
2
are the most abundant gases in the air, M is most likely
N
2
or O
2
, although it can be another gas. The most important reaction creating
ozone is
+ O
2
+MO
3
+M (2.12)
where M, in this case, carries away energy released by the reaction.
Ozone in the stratosphere is destroyed naturally by the photolysis reactions,
O
3
+h O
2
+
O(
1
D) - 310 nm (2.13)
O
3
+h O
2
+
> 310 nm (2.14)

and the two-body reaction
+ O
3
2O
2
(2.15)
Chapman (1930) postulated that the reactions (2.10), (2.12), (2.14), and (2.15),
together with
+ MO
2
+M (2.16)
describe the natural formation and destruction of ozone in the stratosphere. These
reactions make up the Chapman cycle, and they simulate the process fairly well.
Some Chapman reactions are more important than others. The reactions (2.10),
(2.12), and (2.14) affect ozone most signicantly. The non-Chapman reaction,
(2.13), is also important.
The peak ozone density in the atmosphere occurs at about 2532 km
(Chapter 11, Fig. 11.3). The reason is that ozone forms most readily when sufcient
oxygen and ultraviolet radiation are present to produce atomic oxygen through
reactions (2.9) and (2.10). Since oxygen density, like air density, decreases expo-
nentially with increasing altitude, and since ultraviolet radiation intensity increases
with increasing altitude (because radiation attenuates as it passes through the atmo-
sphere), an altitude exists at which the quantities of radiation and oxygen are opti-
mal for producing an ozone maximum. This altitude is between 25 and 32 km. At
higher altitudes, the oxygen density is too lowto produce peak ozone, and at lower
altitudes, the radiation is not intense enough to photodissociate enough oxygen to
produce peak ozone.
Peak stratospheric temperatures occur at the top of the stratosphere, because this
is the altitude at which ozone absorbs the shortest ultraviolet wavelengths reaching
the stratosphere (about 0.175 m). Although the concentration of ozone at the
top of the stratosphere is small, each molecule there can absorb short wavelengths,
increasing the average kinetic energy and temperature of all molecules. In the lower
27
stratosphere, short wavelengths do not penetrate, and temperatures are lower than
in the upper stratosphere.
2.2.2.3 Mesosphere
In the mesosphere, ozone densities are too low for ozone to absorb radiation and
affect temperatures. As such, temperatures decrease with increasing altitude in a
manner similar to in the upper troposphere.
2.2.2.4 Thermosphere
In the thermosphere, temperatures increase with increasing altitude because molec-
ular oxygen and nitrogen there absorb very short wavelengths of solar radiation.
Air in the thermosphere does not feel hot to the skin, because the thermosphere
contains so few gas molecules. But each gas molecule in the thermosphere is highly
energized, so the average temperature is high. Direct sunlight in the thermosphere
is more intense and damaging than at the surface. Because oxygen and nitrogen
absorb the shortest wavelengths of radiation in the thermosphere, such wavelengths
do not penetrate to the mesopause.
2.3 EQUATION OF STATE
The equation of state describes the relationship among pressure, volume, and abso-
lute temperature of a real gas. The ideal gas law describes this relationship for an
ideal gas. An ideal gas is a gas for which the product of the pressure and volume
is proportional to the absolute temperature. A real gas is ideal only to the extent
that intermolecular forces are small, which occurs when pressures are low enough
or temperatures are high enough for the gas to be sufciently dilute. Under typ-
ical atmospheric temperature and pressure conditions, the ideal gas law gives an
error of less than 0.2 percent for dry air and water vapor in comparison with an
expanded equation of state (Pruppacher and Klett 1997). Thus, the ideal gas law
can reasonably approximate the equation of state.
The ideal gas law is expressed as a combination of Boyles law, Charles law,
and Avogadros law. In 1661, Robert Boyle (16271691) found that doubling the
pressure exerted on a gas at constant temperature reduced the volume of the gas
by one-half. This relationship is embodied in Boyles law,
p
1
V
atconstant temperature (2.17)
where p is the pressure exerted on the gas (hPa), and V is the volume enclosed by
the gas (m
3
or cm
3
). Boyles law describes the compressibility of a gas. When high
pressure is exerted on a gas, as in the lower atmosphere, the gas compresses (its
volume decreases) until it exerts an equal pressure on its surroundings. When low
pressure is exerted on a gas, as in the upper atmosphere, the gas expands until it
exerts an equal pressure on its surroundings.
28
2.3 Equation of state
In 1787, Jacques Charles (17461823) found that increasing the absolute tem-
perature of a gas at constant pressure increased the volume of the gas. This rela-
tionship is embodied in Charles law,
V T at constant pressure (2.18)
where T is the temperature of the gas (K). Charles lawstates that, at constant pres-
sure, the volume of a gas must decrease when its temperature decreases. Since gases
change phase to liquids or solids before 0 K, Charles law cannot be extrapolated
to 0 K.
Amedeo Avogadro (17761856) hypothesized that equal volumes of different
gases under the same conditions of temperature and pressure contain the same
number of molecules. In other words, the volume of a gas is proportional to the
number of molecules of gas present and independent of the type of gas. This rela-
tionship is embodied in Avogadros law,
V n at constant pressure and temperature (2.19)
where n is the number of gas moles. The number of molecules in a mole is
constant for all gases and given by Avogadros number, A= 6.0221367
10
23
molec. mol
1
.
Combining Boyles law, Charles law, and Avogadros law gives the ideal gas law
or simplied equation of state as
p =
nR
T
V
=
nA
V
T = Nk
B
T (2.20)
where R
is the universal gas constant (0.0831451 m

3
hPa mol
1
K
1
or 8.31451
10
4
cm
3
hPa mol
1
K
1
), N= nAV is the number concentration of gas molecules
(molecules of gas per cubic meter or cubic centimeter of air), and k
B
= R
A
is Boltzmanns constant in units of 1.380658 10
25
m
3
hPa K
1
molec.
1
or
1.380658 10
19
cm
3
hPa K
1
molec.
1
. Appendix A contains alternative units
for R
and k
B
.
Example 2.3
Calculate the number concentration of air molecules in the atmosphere at
standard sea-level pressure and temperature and at a pressure of 1 hPa.
SOLUTION
At standard sea level, p = 1013 hPa and T = 288 K. Thus, from (2.20), N =
2.55 10
19
molec. cm
3
. From Fig. 2.1(b), p = 1 hPa occurs at 48 km. At this
altitude and pressure, T =270 K, as shown in Fig. 2.4. Under such conditions,
N = 2.68 10
16
molec. cm
3
.
29
Equation (2.20) can be used to relate the partial pressure exerted by a gas to
its number concentration. In 1803, John Dalton (17661844) stated that total
atmospheric pressure equals the sum of the partial pressures of the individual gases
in a mixture. This is Daltons law of partial pressure. The partial pressure exerted
by a gas in a mixture is the pressure the gas exerts if it alone occupies the same
volume as the mixture. Mathematically, the partial pressure of gas q is
p
q
= N
q
k
B
T (2.21)
where N
q
is the number concentration of the gas (molec. cm
3
). Total atmospheric
pressure is
p
a
=
q
p
q
= k
B
T
q
N
q
= N
a
k
B
T (2.22)
where N
a
is the number concentration of the air, determined as the sum of the
number concentrations of individual gases.
Total atmospheric pressure can also be written as p
a
= p
d
+ p
v
, where p
d
is
the partial pressure exerted by dry air, and p
v
is the partial pressure exerted by
water vapor. Similarly, the number concentration of air molecules can be written
as N
a
= N
d
+ N
v
, where N
d
is the number concentration of dry air, and N
v
is the
number concentration of water vapor.
Dry air consists of all gases in the atmosphere, except water vapor. Table 2.1
shows that together, N
2
, O
2
, Ar, and CO
2
constitute 99.996 percent of dry air
by volume. The concentrations of all gases aside from these four can be ignored,
without much loss in accuracy, when dry-air pressure is calculated. This assumption
is convenient in that the concentrations of most trace gases vary in time and space.
Partial pressure is related to the mass density and number concentration of dry
air through the equation of state for dry air,
p
d
=
n
d
R
T
V
=
n
d
m
d
V
m
d
T =
d
R
T =
n
d
A
V
T = N
d
k
B
T (2.23)
where p
d
is dry-air partial pressure (hPa), n
d
is the number of moles of dry air,
m
d
is the molecular weight of dry air,
d
is the mass density of dry air (kg m
3
or
g cm
3
), and R
is the gas constant for dry air. The molecular weight of dry air
is a volume-weighted average of the molecular weights of N
2
, O
2
, Ar, and CO
2
.
The standard value of m
d
is 28.966 g mol
1
. The dry-air mass density, number
concentration, and gas constant are, respectively,
d
=
n
d
m
d
V
N
d
=
n
d
A
V
R
=
R
m
d
(2.24)
where R
has a value of 2.8704 m

3
hPa kg
1
K
1
or 2870.3 cm
3
hPa g
1
K
1
.
Alternative units for R
are given in Appendix A.

30
Example 2.4
When p
d
= 1013 hPa and T = 288 K, the density of dry air from (2.23) is
d
=
1.23 kg m
3
.
The equation of state for water vapor is
p
v
=
n
v
R
T
V
=
n
v
m
v
V
m
v
T =
v
R
v
T =
n
v
A
V
T = N
v
k
B
T (2.25)
where p
v
is the partial pressure exerted by water vapor (hPa), n
v
is the number of
moles of water vapor, m
v
is the molecular weight of water vapor,
v
is the mass
density of water vapor (kg m
3
or g cm
3
), and R
v
is the gas constant for water
vapor (4.6140 m
3
hPa kg
1
K
1
or 4614.0 cm
3
hPa g
1
K
1
). Alternative units for
R
v
are given in Appendix A. The water-vapor mass density, number concentration,
and gas constant are, respectively,
v
=
n
v
m
v
V
N
v
=
n
v
A
V
R
v
=
R
m
v
(2.26)
Example 2.5
When p
v
= 10 hPa and T = 298 K, water vapor density from (2.21) is
v
= 7.27 10
3
kg m
3
.
The equation of state for water vapor can be rewritten in terms of the dry-air
gas constant as
p
v
=
v
R
v
T =
v
R
v
R
T =

v
R
(2.27)
where
=
R
R
v
=
R
m
d
m
v
R
=
m
v
m
d
= 0.622 (2.28)
The number concentration of a gas (molecules per unit volume of air) is an
absolute quantity. The abundance of a gas may also be expressed in terms of a
relative quantity, volume mixing ratio, dened as the number of gas molecules per
molecule of dry air, and expressed for gas q as
q
=
N
q
N
d
=
p
q
p
d
=
n
q
n
d
(2.29)
where N
q
, p
q
, and n
q
are the number concentration, partial pressure, and number
of moles of gas q, respectively. Another relative quantity, mass mixing ratio, is the
31
mass of gas per mass of dry air. The mass mixing ratio of gas q (kilograms of gas
per kilogram of dry air) is
q
=

q
d
=
m
q
N
q
m
d
N
d
=
m
q
p
q
m
d
p
d
=
m
q
n
q
m
d
n
d
=
m
q
m
d
q
(2.30)
where
q
is the mass density (kg m
3
) and m
q
is the molecular weight of gas q
(g mol
1
). Volume and mass mixing ratios may be multiplied by 10
6
and expressed
in parts per million by volume (ppmv) or parts per million by mass (ppmm),
respectively.
Example 2.6
Find the mass mixing ratio, number concentration, and partial pressure of
ozone if its volume mixing ratio in an urban air parcel is
q
= 0.10 ppmv.
Assume T = 288 K and p
d
= 1013 hPa.
SOLUTION
The molecular weight of ozone is m
q
= 48.0 g mol
1
, and the molecular
weight of dry air is m
d
= 28.966 g mol
1
. From (2.30), the mass mixing ratio
of ozone is
q
= 48.0 g mol
1
0.10 ppmv/28.966 g mol
1
= 0.17 ppmm.
From Example 2.3, N
d
=2.55 10
19
molec. cm
3
. Thus, from (2.29), the num-
ber concentration of ozone is N
q
= 0.10 ppmv 10
6
2.55 10
19
molec.
cm
3
= 2.55 10
12
molec. cm
3
. From (2.21), the partial pressure exerted by
ozone is p
q
= 0.000101 hPa.
Combining (2.28) and (2.30) gives the mass mixing ratio of water vapor
(kilograms of water vapor per kilogram of dry air) as
v
=

v
d
=
m
v
p
v
m
d
p
d
=
p
v
p
d
=
p
v
p
a
p
v
=
v
(2.31)
Example 2.7
If the partial pressure exerted by water vapor is p
v
= 10 hPa, and the total air
pressure is p
a
= 1010 hPa, the mass mixing ratio of water vapor from (2.31) is
v
= 0.00622 kg kg
1
.
Water vapor can also be expressed in terms of relative humidity, discussed in
Section 2.5, or specic humidity. The mass of any substance per unit mass of
moist air (dry air plus water vapor) is the moist-air mass mixing ratio (q). Specic
humidity is the moist-air mass mixing ratio of water vapor (mass of water vapor
per unit mass of moist air). An expression for specic humidity (kilograms of water
32
vapor per kilogram of moist air) is
q
v
=

v
a
=

v
d
+
v
=
p
v
R
v
T
p
d
R
T
+
p
v
R
v
T
=
R
R
v
p
v
p
d
+
R
R
v
p
v
=
p
v
p
d
+p
v
(2.32)
where
a
=
d
+
v
is the mass density of moist air. Specic humidity is related to
the mass mixing ratio of water vapor by
q
v
=
v
a
=

v
1 +
v
(2.33)
Example 2.8
If p
v
= 10 hPa and p
a
= 1010 hPa, p
a
= p
d
+ p
v
gives p
d
= 1000 hPa. Under
such conditions, (2.32) gives the specic humidity as q
v
= 0.00618 kg kg
1
.
The equation of state for moist air is the sum of the equations of state for dry
air and water vapor. Thus,
p
a
= p
d
+ p
v
=
d
R
T +
v
R
v
T =
a
R
d
+
v
R
v
R
a
(2.34)
Substituting = R
R
v
,
a
=
d
+
v
, and
v
=
v
d
into (2.34) yields the equa-
tion of state for moist air as
p
a
=
a
R
d
+
v
d
+
v
=
a
R
T
1 +
v
(
d
)
1 +
v
d
=
a
R
T
1 +
v
1 +
v
(2.35)
This equation can be simplied to
p
a
=
a
R
m
T =
a
R
T
v
(2.36)
where
R
m
= R
1 +
v
1 +
v
= R
1 +
1
q
v
= R
(1 +0.608q
v
) (2.37)
is the gas constant for moist air and
T
v
= T
R
m
R
= T
1 +
v
1 +
v
= T
1 +
1
q
v
= T (1 +0.608q
v
) (2.38)
is virtual temperature. This quantity is the temperature of a sample of dry air at the
same density and pressure as a sample of moist air. Since the gas constant for moist
air is larger than that for dry air, moist air has a lower density than does dry air at
the same temperature and pressure. For the dry-air density to equal the moist-air
density at the same pressure, the temperature of the dry air must be higher than
that of the moist air by the factor R
m
R
. The resulting temperature is the virtual

33
temperature, which is always larger than the actual temperature. In (2.38), the
liquid water content is assumed to equal zero.
Equating R
m
= R
m
a
with R
m
from (2.37) and noting that R
= R
m
d
give
the molecular weight of moist air (g mol
1
) as
m
a
=
m
d
1 +0.608q
v
(2.39)
The molecular weight of moist air is less than that of dry air.
Example 2.9
If p
d
= 1013 hPa, p
v
= 10 hPa, and T = 298 K, calculate q
v
, m
a
, R
m
, T
v
, and
a
.
SOLUTION
From (2.32), q
v
= 0.622 10/(1013 + 0.622 10) = 0.0061 kg kg
1
.
From (2.39), m
a
= 28.966/(1 + 0.608 0.0061) = 28.86 g mol
1
.
From (2.37), R
m
= 2.8704 (1 + 0.608 0.0061) = 2.8811 m
3
hPa kg
1
K
1
.
From (2.38), T
v
= 298 (1 + 0.608 0.0061) = 299.1 K.
From (2.36),
a
= p
a
/(R
m
T) = 1023/(2.8811 298) = 1.19 kg m
3
.
2.4 CHANGES OF PRESSURE WITH ALTITUDE
The variation of pressure with altitude in the atmosphere can be estimated several
ways. The rst is by considering the equation for hydrostatic air pressure given in
(2.1). The differential form of the equation, called the hydrostatic equation, is
dp
a
=
a
g dz (2.40)
where air pressure, air density and gravity are functions of altitude (z), but paren-
theses have been removed. The negative sign arises because pressure decreases with
increasing altitude. The hydrostatic equation assumes that the downward force of
gravity per unit volume of air (
a
g) exactly balances an upward pressure gra-
dient force per unit volume (p
a
z). Figure 2.1(b) shows that pressure decreases
exponentially with increasing altitude, giving rise to an upward pressure gradient
force (a force that moves air from high to low pressure Section 4.2.4). If gravity
did not exist, the upward pressure gradient would accelerate air to space.
Pressure at a given altitude can be estimated from the hydrostatic equation with
p
a.k
p
a.k+1
a.k+1
g
k+1
(z
k
z
k+1
) (2.41)
where p
a.k
is the pressure at any upper altitude z
k
, and p
a.k+1
,
a.k+1
, and g
k+1
are pressure, density, and gravity, respectively, at any lower altitude, z
k+1
. If the
pressure is known at the surface, and a vertical density and gravity prole are
34
2.4 Changes of pressure with altitude
known, (2.41) can be used to estimate the pressure at each altitude above the
surface, as illustrated in Example 2.10. Appendix Table B.1 gives vertical proles
of density and gravity in the atmosphere.
Example 2.10
If pressure, density, and gravity at sea level are p
a.k+1
= 1013.25 hPa,
a.k+1
=
1.225 kg m
3
, and g
k+1
= 9.8072 ms
2
, respectively, estimate the pressure at
100-m altitude.
SOLUTION
From(2.41), p
a.100 m
= 1013.25 hPa 1.225
kg
m
3
9.8072
m
s
2
(100 0m)
hPa ms
2
100 kg
=
1001.24 hPa, which compares well with the standard atmosphere pressure at
100 m given in Appendix Table B.1. This result also suggests that, near the
surface, pressure decreases approximately 1 hPa for every 10 m increase in
altitude.
Altitude versus pressure can also be determined from a pressure altimeter,
which measures the pressure at an unknown altitude with an aneroid barometer
(Section 1.1). From the barometric pressure, altitude is calculated under standard
atmospheric conditions. Under such conditions, sea-level air pressure, sea-level air
temperature, and the environmental lapse rate (average negative change in air tem-
perature with altitude in the free troposphere) are p
a.s
=1013.25 hPa, T
a.s
=288 K,
and I
s
= Tz = +6.5 K km
1
, respectively (List 1984).
The equation for altitude in a standard atmosphere is derived by substituting
p
a
=
a
R
m
T into the hydrostatic equation from (2.40). The result is
p
a
z
=
p
a
R
m
T
g (2.42)
Substituting T = T
a.s
I
s
z, where z = 0 km corresponds to sea level, into (2.42),
rearranging, then integrating from p
a.s
to p
a
and 0 to z yields
ln
p
a
p
a.s
=
g
I
s
R
m
ln
T
a.s
I
s
z
T
a.s
(2.43)
Rearranging again gives altitude as a function of pressure in a standard troposphere
as
z =
T
a.s
I
s
p
a
p
a.s
I
s
R
m
g
(2.44)
Temperature variations with altitude and sea-level pressure in the real atmosphere
usually differ from those in the standard atmosphere. Empirical and tabulated
expressions correcting for the differences are available (List 1984). Since the
35
corrections are not always accurate, airplanes rarely use pressure altimeters to
measure altitude. Instead, they use radar altimeters, which measure altitude with
radio waves.
Example 2.11
If a pressure altimeter reads p
a
= 850 hPa and the air is dry, nd the standard-
atmosphere altitude.
SOLUTION
Because the air is dry, R
m
= R
= 287.04 m
2
s
2
K
1
. Thus, I
s
R
m
g = 0.1902,
and from (2.44), z = 1.45 km.
Athird way to estimate pressure versus altitude is with the scale-height equation.
From the equation of state for moist air, air density is
a
=
p
a
R
T
v
=
m
d
R
p
a
T
v
=
p
a
T
v
A
R
m
d
A

p
a
T
v
1
k
B

M=
p
a

M
k
B
T
v
(2.45)
where

M m
d
Awas previously dened as the average mass of one air molecule.
Equation (2.45) can be combined with the hydrostatic equation to give
dp
a
p
a
=
Mg
k
B
T
v
dz =
dz
H
(2.46)
where
H =
k
B
T
v
Mg
(2.47)
is the scale height of the atmosphere at a given virtual temperature. The scale height
is the height above a reference height at which pressure decreases to 1/e of its value
at the reference height. Since temperature varies with altitude in the atmosphere,
scale height also varies with altitude.
Example 2.12
Determine the scale height at T
v
= 298 K and p
a
= 1013.25 hPa. Assume dry
air.
SOLUTION
From Appendix A and Appendix Table B.1,

M = 4.8096 10
26
kg, g =
9.8072 ms
2
, and k
B
= 1.380658 10
23
kg m
2
s
2
K
1
molec.
1
at the given
pressure. Thus, from (2.47), the scale height is H = 8.72 km.
36
2.5 Water in the atmosphere
Integrating (2.46) at constant temperature gives pressure as a function of
altitude as
p
a
= p
a.ref
e
(zz
ref
)H
(2.48)
where p
a.ref
is pressure at a reference level, z =z
ref
. In a model with several vertically
stacked layers, each with a known average temperature, pressure at the top of each
layer can be calculated from (2.48) assuming that the reference pressure at the base
of the layer is the pressure calculated for the top of the next-lowest layer. At the
ground, the pressure is the surface pressure, which must be known.
2.5 WATER IN THE ATMOSPHERE
Water appears in three states gas, liquid, and solid. Sources and sinks of water
vapor in the atmosphere were discussed in Section 2.1.2.1. Sources of liquid water
include sea spray emission, volcanos, combustion, condensation of water vapor,
and melting of ice crystals. Sinks of liquid water include evaporation, freezing, and
sedimentation to the surface of aerosol particles, fog drops, drizzle, and raindrops.
Ice in the atmosphere forms from freezing of liquid water and solid deposition of
water vapor. Losses of ice occur by sublimation to the gas phase, melting to the
liquid phase, and sedimentation to the surface.
2.5.1 Types of energy
When water changes state, it releases or absorbs energy. Energy is the capacity of a
physical system to do work on matter. Matter is mass that exists as a solid, liquid,
or gas. Energy takes many forms.
Kinetic energy is the energy within a body due to its motion and equals one-half
the mass of the body multiplied by its speed squared. The faster a body moves,
the greater its kinetic energy. To change the speed and kinetic energy of a body in
motion, mechanical work must be done.
Potential energy is the energy of matter that arises due to its position, rather than
its motion. Potential energy represents the amount of work that a body can do. A
coiled spring, charged battery, and chemical reactant have potential energy. When
an object is raised vertically, it possesses gravitational potential energy because it
can potentially do work by sinking.
Internal energy is the kinetic and/or potential energy of molecules within an
object, but does not include the kinetic and/or potential energy of the object as a
whole.
Work is the energy added to a body by the application of a force that moves the
body in the direction of the force.
Electromagnetic (radiant) energy is the energy transferred by electromagnetic
waves that originate from bodies with temperatures above 0 K.
37
Water drop
Freezing
Melting
Condensation
Evaporation
Sublimation
Deposition
Ice crystal
W
a
t
e
r

v
a
p
o
r
Figure 2.6 Phase changes of water. Freezing (melting) at 0
Creleases (absorbs)
333.5 J g
1
, deposition (sublimation) at 0

C releases (absorbs) 2835 J g
1
,
and condensation (evaporation) releases (absorbs) 2510 J g
1
at 0

C and
2259 J g
1
at 100

C.
Heat, itself, is not a form of energy. However, heat transfer is a term used to
describe the energy transfer between two bodies that occurs, for example, when
their internal energies (or temperatures) differ. Heat capacity (e.g., specic heat
capacity) is the energy required to change the temperature of a given quantity of a
substance 1

C. Finally, heat release (absorption) occurs when a substance releases
(absorbs) energy to (from) the surrounding environment upon a change of state.
Below, heat release/absorption is discussed.
2.5.2 Latent heat
During condensation, freezing, and solid deposition of a substance, energy is
released. During evaporation, melting, and sublimation, energy is absorbed. The
stored energy released or the energy absorbed during such processes is called latent
heat. Latent heat absorbed (released) during evaporation (condensation) is latent
heat of evaporation. It varies with temperature as
dL
e
dT
= c
p.V
c
W
(2.49)
where c
p.V
is the specic heat of water vapor at constant pressure, and c
W
is the
specic heat of liquid water. The latent heat absorbed (released) during sublimation
(deposition) is the latent heat of sublimation. The latent heat absorbed (released)
during melting (freezing) is the latent heat of melting. The latent heats of sublima-
tion and melting vary with temperature according to
dL
s
dT
= c
p.V
c
I
dL
m
dT
= c
W
c
I
(2.50)
respectively, where c
I
is the specic heat of ice. Figure 2.6 gives the quantity of
energy absorbed or released during phase changes of water.
38
4000
4500
5000
5500
6000
40 30 20 10 0 10 20 30 40
(
J

k
g
1

K
1
)
Temperature (C)
c
W
Figure 2.7 Variation of specic heat of
liquid water with temperature, from
(2.51).
The specic heats of liquid water, ice, and water vapor vary with tempera-
ture. Polynomial expressions for the specic heat of liquid water (J kg
1
K
1
)
are
c
W
=
4187.9 11.319T
c
0.097215T
2
c
+0.018317T
3
c
+0.0011354T
4
c
37 T
c
- 0
4175.2 +0.01297(T
c
35)
2
+1.5899 10
5
(T
c
35)
4
0 T
c
- 35
(2.51)
(Osborne et al. 1939; Angell et al. 1982; Pruppacher and Klett 1997), where T
c
is temperature in degrees Celsius. Figure 2.7 shows c
W
versus temperature from
(2.51). The variation of c
W
below 0

C is large, but that above 0

C is small. An
expression for the specic heat of ice is
c
I
= 2104.6 +7.322T
c
40 T
c
- 0
C (2.52)
(Giauque and Stout 1936; Flubacher et al. 1960; Pruppacher and Klett 1997).
The specic heat of water vapor at constant pressure also varies slightly with
temperature. At 298.15 K, c
p.V
1865.1 (Lide 1993). At 303.15 K, it is 2 percent
larger than at 243.15 K (Rogers and Yau 1989).
Because changes in c
W
and c
p.V
are small for temperatures above 0

C, these
parameters may be held constant when the variation of the latent heat of evapora-
tion with temperature is calculated. Integrating (2.49) with constant specic heat
gives
L
e
= L
e.0
(c
W
c
p.V
) (T T
0
) (2.53)
where L
e.0
is the latent heat of evaporation at temperature T
0
(K). An empiri-
cal expression for the latent heat of evaporation (J kg
1
) is (Bolton 1980; List
1984)
L
e
2.501 10
6
2370T
c
(2.54)
39
Example 2.13
Equation (2.54) predicts that, at 0

C, 2501 J (about 600 cal) is required to
evaporate 1 g of liquid water. At 100

C, 2264 J (about 540 cal) is required to
evaporate 1 g of liquid water.
An empirical expression for the latent heat of melting (J kg
1
), valid for temper-
atures below 0

C, is
L
m
3.3358 10
5
+ T
c
(2030 10.46T
c
) (2.55)
(List 1984). While pure water at standard pressure always melts at temperatures
above 0

C, it may or may not freeze at temperatures below 0

C. Water that
remains liquid below 0

C is supercooled liquid water.
Example 2.14
Equation (2.55) predicts that, when 1 g of liquid water freezes, 333.5 J (about
80 cal) is released at 0

C and 312.2 J (74.6 cal) is released at 10

C.
The latent heat of sublimation (J kg
1
) is the sum of the latent heats of evapo-
ration and melting. Thus,
L
s
= L
e
+ L
m
2.83458 10
6
T
c
(340 +10.46T
c
) (2.56)
2.5.3 ClausiusClapeyron equation
The rates of formation and growth of liquid water drops and ice crystals depend
on several factors, described in Chapters 16 and 18. One important parameter,
derived here, is the saturation vapor pressure (SVP, also called the equilibrium
vapor pressure or surface vapor pressure). This parameter is the partial pressure of
a gas over a particle surface at a specic temperature when the gas is in equilibrium
with its liquid or solid phase on the surface.
Asimplistic way of looking as SVP is to viewit as the maximumamount of vapor
the air can sustain without the vapor condensing as a liquid or depositing as ice on
a surface. When the air is cold, SVPs are lower than when the air is warm. At low
temperatures, liquid-water molecules on a particle surface have little kinetic energy
and cannot easily break free (evaporate) from the surface. Thus, in equilibrium, air
just above a surface contains few vapor molecules and exerts a low SVP. When the
air is warm, liquid-water molecules have higher kinetic energies, are more agitated,
and break loose more readily from a surface. The resulting evaporation increases
the vapor content of the air, and thus the SVP over the particle surface.
40
The temperature variation of the saturation vapor pressure of water over a liquid
surface ( p
v.s
, hPa) is approximated with the ClausiusClapeyron equation,
dp
v.s
dT
=

v.s
T
L
e
(2.57)
where L
e
is the latent heat of evaporation of water (J kg
1
), and
v.s
is the saturation
mass density of water vapor (kg m
3
). Combining
v.s
= p
v.s
R
v
T from(2.25) with
(2.57) gives
dp
v.s
dT
=
L
e
p
v.s
R
v
T
2
(2.58)
Substituting L
e
from (2.54) into (2.58) and rearranging give
dp
v.s
p
v.s
=
1
R
v
A
h
T
2

B
h
T
dT (2.59)
where A
h
= 3.14839 10
6
J kg
1
and B
h
= 2370 J kg
1
K
1
. Integrating (2.59)
from p
v.s.0
to p
v.s
and T
0
to T, where p
v.s.0
is a known saturation vapor pressure
at T
0
, gives the saturation vapor pressure of water over a liquid surface as
p
v.s
= p
v.s.0
exp
A
h
R
v
1
T
0
1
T
+
B
h
R
v
ln
T
0
T
(2.60)
At T = T
0
= 273.15 K, p
v.s.0
= 6.112 hPa. Substituting these values, R
v
=
461.91 J kg
1
K
1
, A
h
, and B
h
into (2.60) gives
p
v.s
= 6.112 exp
6816
1
273.15

1
T
+5.1309 ln
273.15
T
(2.61)
where T is in kelvin and p
v.s
is in hPa.
Example 2.15
From (2.60), p
v.s
= 1.26 hPa at T = 253.15 K (20

C) and p
v.s
= 31.60 hPa at
T = 298.15 K (25

C).
An empirical parameterization of the saturation vapor pressure of water over a
liquid surface is
p
v.s
= 6.112 exp
17.67T
c
T
c
+243.5
(2.62)
(Bolton 1980), where T
c
is in degrees Celsius and p
v.s
is in hPa. The t is valid
for 35 - T
c
- 35

C. Equations (2.61) and (2.62) are saturation vapor pressures
over at, dilute liquid surfaces. Saturation vapor pressures are affected by surface
41
0
20
40
60
80
100
120
20 10 0 10 20 30 40 50
V
a
p
o
r

p
r
e
s
s
u
r
e

(
h
P
a
)
Temperature (C)
Over liquid
water
(a)
0
1
2
3
4
5
6
7
8
50 40 30 20 10
V
a
p
o
r

p
r
e
s
s
u
r
e

(
h
P
a
)
Temperature (C)
Over liquid
water
Over ice
0
(b)
Figure 2.8 Saturation vapor pressure over (a) liquid water and (b) liquid
water and ice, versus temperature.
curvature, the presence of solutes in water, and the rate of drop heating or cooling.
These effects are discussed in Chapter 16.
Example 2.16
At T = 253.15 K (20

C), the saturation vapor pressure from (2.62) is p
v.s
=
1.26 hPa. At T = 298.15 K (25

C), p
v.s
= 31.67 hPa. A comparison of these
results with those from the previous examples shows that (2.61) and (2.62)
give almost identical estimates for p
v.s
.
Figure 2.8(a) shows saturation vapor pressure of water over a liquid surface ver-
sus temperature, obtained from (2.62). The gure indicates that saturation vapor
pressure increases with increasing temperature. At 0

C, p
v.s
= 6.1 hPa, which is
equivalent to 0.6 percent of the total sea-level air pressure (1013 hPa). At 30

C,
p
v.s
=42.5 hPa (4.2 percent of the total sea-level air pressure). Since the partial pres-
sure of water vapor can rarely exceed its saturation vapor pressure, the maximum
partial pressure of air varies from-1 percent to 4 percent of total atmospheric air
pressure. Near the poles, where temperatures are low, saturation vapor pressures
and partial pressures approach zero. Near the Equator, where temperatures are
high, saturation vapor pressures and partial pressures can increase to 4 percent or
more of the total air pressure.
The saturation vapor pressure of water over an ice surface is lower than that
over a liquid surface at the same subfreezing temperature because ice molecules
require more energy to sublimate than liquid molecules require to evaporate at
the same temperature. The ClausiusClapeyron equation for the saturation vapor
pressure of water over ice ( p
v.I
, hPa) is rewritten from (2.58) as
dp
v.I
dT
=
L
s
p
v.I
R
v
T
2
(2.63)
42
Substituting the latent heat of sublimation from (2.56) into (2.63) and integrating
give the saturation vapor pressure of water over ice as
p
v.I
= 6.112 exp
4648
1
273.15

1
T
11.64 ln
273.15
T
+ 0.02265(273.15 T)
(2.64)
where T 273.15 K is in kelvin and p
v.I
is in hPa. At T = 273.15 K, the sat-
uration vapor pressure over ice equals that over liquid water ( p
v.I.0
= 6.112
hPa). Figure 2.8(b) shows that, at subfreezing temperatures, p
v.I
- p
v.s
. Above
273.15 K (0

C), ice surfaces do not exist. An alternative parameterization, valid
from 223.15 K to 273.15 K is
p
v.I
= 6.1064 exp
21.88(T 273.15)
T 7.65
(2.65)
(Pruppacher and Klett 1997), where T 273.15 K is in kelvin and p
v.I
is in hPa.
Example 2.17
At T = 253.15 K (20

C), (2.64) gives p
v.I
= 1.034 hPa, which is less than
p
v.s
= 1.26 hPa at the same temperature. Thus, the saturation vapor pressure
of water over liquid is greater than that over ice.
2.5.4 Condensation and deposition
Saturation vapor pressures are critical for determining the extent of liquid drop
and ice crystal formation. When the air temperature is above the freezing tem-
perature of water (273.15 K) and the partial pressure of water vapor is greater
than the saturation vapor pressure of water over a liquid surface ( p
v
> p
v.s
), water
vapor generally condenses as a liquid. Liquid water evaporates when p
v
- p
v.s
.
Figures 2.9(a) and (b) illustrate growth and evaporation.
If the air temperature falls below freezing and p
v
> p
v.I
, water vapor deposits
as ice. When p
v
- p
v.I
, ice sublimates to water vapor. If liquid water and ice coex-
ist in the same parcel of air and p
v.I
- p
v.s
- p
v
, deposition onto ice particles
is favored over condensation onto liquid water drops because growth is driven
by the difference between partial pressure and saturation vapor pressure, and
p
v
p
v.I
> p
v
p
v.s
.
When p
v.I
- p
v
- p
v.s
, water evaporates from liquid surfaces and deposits onto
ice surfaces, as shown in Fig. 2.10. This is the central assumption behind the
WegenerBergeronFindeisen (Bergeron) process of ice crystal growth in cold
clouds. In such clouds, supercooled liquid water cloud droplets coexist with ice
crystals. When the ratio of liquid cloud drops to ice crystals is less than 100 000:1,
each ice crystal receives water that evaporates from less than 100 000 cloud drops,
and ice crystals do not grow large or heavy enough to fall from their cloud. When
the ratio is greater than 1 000 000:1, the relatively few ice crystals present grow
43
Condensation
Particle
Gas
(a) Evaporation
Particle
Gas
(b)
Figure 2.9 (a) Condensation occurs when the partial pressure of
a gas exceeds its saturation vapor pressure over a particle surface.
(b) Evaporation occurs when the saturation vapor pressure exceeds
the partial pressure of the gas. The schematics are not to scale.
Ice crystal
Water vapor
Water droplet
Figure 2.10 Bergeron process. When p
v.I
- p
v
- p
v.s
at a subfreezing
temperature, liquid evaporates, and the resulting gas diffuses to and
deposits onto ice crystals.
large and fall quickly from their cloud before much liquid water evaporates, leav-
ing the cloud with lots of liquid drops too small to fall from the cloud. When the
ratio is between 100 000:1 and 1 000 000:1, each ice crystal receives the liquid
water from 100 000 to 1 000 000 droplets. These ice crystals fall from the cloud,
maximizing precipitation.
Figure 2.8(b) shows that the greatest difference between the saturation vapor
pressures over liquid water and ice occurs at T
c
=15
C. This is the temperature at

which ice crystals growthe fastest. Dendrites, which are snowake-shaped crystals,
form most favorably at this temperature.
The ratio of the water vapor partial pressure to its saturation vapor pressure over
a liquid surface determines whether the vapor condenses. This ratio is embodied
in the relative humidity, dened as
f
r
= 100%

v
v.s
= 100%
p
v
( p
a
p
v.s
)
p
v.s
( p
a
p
v
)
100%
p
v
p
v.s
(2.66)
44
by the World Meteorological Organization (WMO), where
v
is the mass mixing
ratio of water vapor in the air from (2.31), and
v.s
=
p
v.s
p
a
p
v.s
p
v.s
p
d
(2.67)
is the saturation mass mixing ratio of water vapor over a liquid surface, found
by substituting p
v.s
for p
v
and
v.s
for
v
in (2.31). The saturation mass mixing
ratio in the equation for the relative humidity is always taken with respect to liquid
water, even if T 273.15 K (List 1984). Prior to the WMO denition, the relative
humidity was dened exactly as f
r
= 100% p
v
p
v.s
.
Example 2.18
If T = 288 K and p
v
= 12 hPa, what is the relative humidity?
SOLUTION
From (2.62), p
v.s
= 17.04 hPa. From (2.66), f
r
= 100 percent 12 hPa / 17.04
hPa = 70.4 percent.
Equation (2.66) implies that, if the relative humidity exceeds 100 percent and
T >273.15 K, water vapor condenses. If the relative humidity exceeds 100 percent
and T 273.15 K, water vapor may condense as a liquid or deposit as ice.
Another parameter used to predict when bulk condensation occurs is the dew
point (T
D
), which is the temperature to which air must be cooled, at constant
water vapor partial pressure and air pressure, to reach saturation with respect
to liquid water. Similarly, the frost point is the temperature to which air must
be cooled, at constant water vapor partial pressure and air pressure, to reach
saturation with respect to ice. If the air temperature drops below the dew point,
the relative humidity increases above 100 percent and condensation occurs. If the
dew point is known, the partial pressure of water can be obtained from Fig. 2.8. If
the ambient temperature is known, the saturation vapor pressure can be obtained
from the same gure.
Example 2.19
If T
D
=20
CandT=30
C, estimate the partial pressure of water, the saturation

vapor pressure of water, and the relative humidity from Fig. 2.8.
SOLUTION
From the gure, p
v
23.4 hPa and p
v.s
42.5 hPa. Thus, f
r
55 percent.
An equation for the dew point can be derived from any equation for saturation
vapor pressure by substituting p
v
for p
v.s
and solving for the temperature. Applying
45
260 270 280 290 300 310
700
750
800
850
900
950
1000
Temperature (K)
P
r
e
s
s
u
r
e

(
h
P
a
)
3:30 a.m.
Temperature
Dew point
(a)
260 270 280 290 300 310
700
750
800
850
900
950
1000
Temperature (K)
P
r
e
s
s
u
r
e

(
h
P
a
)
Temperature
Dew point
3:30 p.m.
(b)
Figure 2.11 Observed vertical proles of temperature and dew point at (a) 3:30
a.m. and (b) 3:30 p.m. on August 27, 1987 at Riverside, California. The air was
nearly saturated near the surface in the morning.
this technique to (2.62) and converting the result to absolute temperature give the
dew point as
T
D
=
4880.357 29.66 ln p
v
19.48 ln p
v
=
4880.357 29.66 ln(
v
p
d
)
19.48 ln(
v
p
d
)
(2.68)
where
v
= p
v
p
d
from (2.31), T is in kelvin, and p
v
is in hPa.
Example 2.20
Calculate the dew point when p
v
= 12 hPa.
SOLUTION
From (2.68), T
D
= 282.8 K. Thus, if unsaturated air at 288 K cools to 282.8 K
when p
v
= 12 hPa, condensation occurs.
When the ambient temperature is close to the dew point (T T
D
), the relative
humidity is high. When the ambient temperature and dew point are far apart, the
relative humidity is low. Figures 2.11(a) and (b) show observed vertical proles of
temperature and dew point at Riverside, California, in the morning and afternoon,
respectively, of August 27, 1987. In the morning, near the ground, the dew point
and air temperature were close to each other, indicating the air was nearly saturated,
the relative humidity was high, and a fog was almost present. Above 950 hPa (about
500 m above sea level) in the morning, the dew point and air temperature were far
apart, indicating the relative humidity was low and the air was unsaturated. In the
afternoon, when air near the surface was warm, the dew point and air temperature
were also far apart, and the relative humidity was low.
46
2.6 First law of thermodynamics
2.6 FIRST LAW OF THERMODYNAMICS
As applied to the atmosphere, the rst law of thermodynamics relates the change
in temperature of a parcel of air to energy transfer between the parcel and its envi-
ronment and work done by or on the parcel. The rst law is used to derive the
thermodynamic energy equation, which gives the time-rate-of-change of tempera-
ture in the atmosphere due to energy transfer and work. The rst law is also used
to derive analytical expressions for atmospheric stability.
The rst law of thermodynamics as applied to the atmosphere is
dQ
= dU
+dW
(2.69)
where dQ
is called the diabatic heating term, which is the energy (J) trans-
ferred between an air parcel and its environment, dU
is the change in internal

energy (J) of the parcel, and dW
is the work (J) done by or on the parcel. When

dQ
> 0, energy is transferred to the parcel from the environment, and the process
is endothermic. When dQ
- 0, energy is transferred to the environment from the

parcel, and the process is exothermic. When dW
> 0, work is done by the parcel.

When dW
- 0, work is done on the parcel. Equation (2.69) states that if energy

is added to an air parcel, some of it is used to change the internal energy (and
temperature) of the parcel, and the rest is used by the parcel to do work. The equa-
tion also states that changes in the internal energy (temperature) of a parcel result
from energy transfer or work. Internal energy changes resulting from energy trans-
fer are diabatic while those resulting from work are adiabatic. Diabatic sources
or sinks of energy include conduction, radiative cooling/heating, and latent heat
release/absorption. Substituting
dQ =
dQ
M
a
dU =
dU
M
a
dW=
dW
M
a
(2.70)
where M
a
= M
d
+ M
v
is the mass of a parcel of air (kg) consisting of dry air mass
M
d
and water vapor mass M
v
, into (2.69) gives the rst law in terms of energy per
unit mass of air (J kg
1
) as
dQ = dU +dW (2.71)
Terms in this equation are discussed below.
When a gas expands, work is done by the gas. When air expands, work done
by the air is dW
= p
a
dV, and work done per unit mass of air is
dW=
dW
M
a
=
p
a
dV
M
a
= p
a
d
a
(2.72)
In this equation, dV is the change in volume of the air, and
a
=
V
M
a
=
1
a
(2.73)
is the specic volume of air. Air expands when it rises to lower pressure. In such
cases, work is done by the air, and dV > 0. When a parcel of air sinks to higher
pressure, the parcel compresses, work is done on the air in the parcel, and dV - 0.
47
Energy transfer between a parcel and its environment occurs when dQ = 0. In the
atmosphere, major sources (sinks) of external energy are radiative heating (cooling),
condensation (evaporation), deposition (sublimation), and freezing (melting).
The change in internal energy of the air is its change in temperature multiplied
by the energy required to change its temperature one degree Celsius (1 K) without
affecting the volume or work done by or on the air. In other words,
dU =
Q
T
a
dT = c
v.m
dT (2.74)
where c
v.m
= (QT)
a
is the specic heat of moist air at constant volume. It is the
energy required to raise the temperature of 1 g of air 1 K at constant volume and
varies with water-vapor mass mixing ratio. An expression for c
v.m
can be derived
by noting that, at constant volume,
(M
d
+ M
v
)dQ =(M
d
c
v.d
+ M
v
c
v.V
)dT (2.75)
where c
v.d
= 717.63 J kg
1
K
1
at 298 K is the specic heat of dry air at constant
volume and c
v.V
=1403.2 J kg
1
K
1
is the specic heat of water vapor at constant
volume. The specic heat of dry air at constant volume decreases by less than 0.2
percent down to 200 K. Dividing (2.75) through by (M
d
+ M
v
) dT gives
c
v.m
=
Q
T
a
=
M
d
c
v.d
+ M
v
c
v.V
M
d
+ M
v
=
c
v.d
+c
v.V
v
1 +
v
= c
v.d
(1 +0.955q
v
) (2.76)
where q
v
=
v
(1 +
v
) from (2.33) and c
v.V
c
v.d
1 = 0.955.
Substituting (2.74) and (2.72) into (2.71) gives the rst law of thermodynamics
for the atmosphere as
dQ = c
v.m
dT + p
a
d
a
(2.77)
Combining the equation of state, p
a
=
a
R
m
T, with
a
= 1
a
gives p
a
a
= R
m
T,
which can be differentiated to give
p
a
d
a
+
a
dp
a
= R
m
dT (2.78)
Combining (2.78) with (2.77) yields another form of the rst law as
dQ = c
p.m
dT
a
dp
a
(2.79)
where
c
p.m
=
dQ
dT
p
a
=
M
d
c
p.d
+ M
v
c
p.V
M
d
+ M
v
=
c
p.d
+c
p.V
v
1 +
v
= c
p.d
(1 +0.856q
v
)
c
p.d
(1 +0.859
v
) (2.80)
48
is the specic heat of moist air at constant pressure. In this equation, c
p.V
c
p.d
1
= 0.856, where c
p.d
= 1004.67 J kg
1
K
1
at 298 K is the specic heat of dry air
at constant pressure, and c
p.V
= 1865.1 J kg
1
K
1
is the specic heat of water
vapor at constant pressure. Like c
v.d
, c
p.d
and c
p.V
vary slightly with temperature.
The specic heat of moist air at constant pressure is the energy required to increase
the temperature of 1 g of air 1 K without affecting air pressure.
Differentiating (2.77) at constant pressure gives another expression for c
p.m
as
c
p.m
=
dQ
dT
p
a
= c
v.m
+ p
a
d
a
dT
p
a
= c
v.m
+ p
a
d
dT
R
m
T
p
a
p
a
= c
v.m
+ R
m
(2.81)
When the air is dry, this expression simplies to c
p.d
= c
v.d
+ R
. Substituting c
v.m
=
c
v.d
(1 +0.955q
v
) from (2.76) and R
m
= R
(1 +0.608q
v
) from (2.37) into (2.81)
gives the empirical expressions in (2.80).
The rst law of thermodynamics can be approximated in terms of virtual tem-
perature instead of temperature with
dQ =
1 +0.856q
v
1 +0.608q
v
c
p.d
dT
v
a
dp
a
c
p.d
dT
v
a
dp
a
(2.82)
which was derived by substituting c
p.m
= c
p.d
(1 +0.856q
v
) from (2.80) and T =
T
v
(1 +0.608q
v
) from(2.38) into (2.79). An advantage of (2.82) is that the specic
heat is in terms of dry instead of moist air. The maximum energy error in (2.82)
due to neglecting the empirical water vapor terms is 1 percent. The average error
is 0.2 percent.
2.6.1 Applications of the rst law of thermodynamics
Here, the rst law of thermodynamics is modied for four special cases. First, for
an isobaric process (dp
a
= 0), the rst law simplies from (2.79) to
dQ = c
p.m
dT =
c
p.m
c
v.m
dU (2.83)
Second, for an isothermal process (dT = 0), (2.79) becomes
dQ =
a
dp
a
= p
a
d
a
= dW (2.84)
Third, for an isochoric process (d
a
= 0), the rst law simplies from (2.77) to
dQ = c
v.m
dT = dU (2.85)
A fourth case is for an adiabatic process. Under adiabatic conditions, no energy
is transferred to or from a parcel of air (dQ = 0). Instead, a parcels temperature
changes only when the parcel expands or contracts as it ascends or descends,
respectively. When a parcel rises, it encounters lower pressure and expands. As
it expands, the kinetic energy of the air molecules within the parcel is converted
to work used to expand the air. Since temperature decreases proportionally to
kinetic energy from (2.4), a rising, expanding parcel of air cools under adiabatic
49
conditions. The expansion of an air parcel in the absence of diabatic sources or
sinks of energy is called an adiabatic expansion.
When a parcel sinks, it compresses and warms. When no energy transfer is
considered, the compression is called an adiabatic compression. Under adiabatic
conditions, the rst law can be rewritten from (2.77), (2.79), and (2.82), respec-
tively, to
c
v.m
dT = p
a
d
a
(2.86)
c
p.m
dT =
a
dp
a
(2.87)
c
p.d
dT
v

a
dp
a
(2.88)
2.6.1.1 Dry adiabatic lapse rate
When an air parcel rises and cools, and no condensation occurs, the parcels rate of
cooling with height is approximately 9.8 K km
1
, which is the dry (unsaturated)
adiabatic lapse rate. A lapse rate is the negative change in temperature with height.
A positive lapse rate indicates that temperature decreases with increasing height.
The dry adiabatic lapse rate can be derived fromthe hydrostatic equation and the
adiabatic form of the rst law of thermodynamics. Taking the negative differential
of (2.88) with respect to altitude, substituting the hydrostatic equation from(2.40),
and substituting
a
= 1
a
give the dry adiabatic lapse rate in terms of virtual
temperature as
I
d
=
T
v
z

a
c
p.d
p
a
z
=

a
c
p.d
a
g =
g
c
p.d
= +9.8 K km
1
(2.89)
where the subscript d indicates that the change is dry (unsaturated) adiabatic.
Equation (2.89) states that the virtual temperature of an unsaturated air parcel
cools 9.8 K for every kilometer it ascends in the atmosphere under dry adiabatic
conditions. From(2.87), the dry adiabatic lapse rate in terms of actual temperature
can be written as
I
d.m
=
T
z
d
=
g
c
p.m
=
g
c
p.d
1 +
v
1 +c
p.V
v
c
p.d
(2.90)
where c
p.m
, obtained from (2.80), varies with water vapor content. The advantage
of (2.89) is that the right side of the equation does not depend on water vapor
content.
2.6.1.2 Potential temperature
A parameter used regularly in atmospheric analysis and modeling is potential tem-
perature. Potential temperature is the temperature an unsaturated air parcel attains
if it is brought adiabatically from its altitude down to a pressure of 1000 hPa.
Potential temperature is derived by rst substituting
a
= R
m
Tp
a
into (2.87),
giving
dT
T
=
R
m
c
p.m
dp
a
p
a
(2.91)
50
Integrating (2.91) from T
0
to T and p
a.0
to p
a
yields Poissons equation,
T = T
0
p
a
p
a.0
R
m
c
p.m
= T
0
p
a
p
a.0
(1+0.608q
v)
c
p.d
(1+0.856q
v)
T
0
p
a
p
a.0
(10.251q
v
)
(2.92)
where
=
R
c
p.d
=
c
p.d
c
v.d
c
p.d
= 0.286 (2.93)
When p
a.0
= 1000 hPa, T
0
is called the potential temperature of moist air (
p.m
),
and (2.92) becomes
p.m
= T
1000 hPa
p
a
(10.251q
v
)
(2.94)
In the absence of water vapor, q
v
= 0 and p
a
= p
d
. In such a case, (2.94) simplies
to
p
= T
1000 hPa
p
d
(2.95)
where
p
is the potential temperature of dry air. Since q
v
is usually smaller than
0.03 kg kg
1
, neglecting q
v
in (2.94) causes an error in of less than 0.75 percent.
Thus, for simplicity, (2.95) is usually used instead of (2.94) for dening potential
temperature, even when water vapor is present. Potential temperature is conserved
(stays constant) if an unsaturated air parcel is displaced adiabatically.
Figure 2.12 shows vertical proles of potential temperature in the morning and
afternoon at Riverside, California on August 27, 1987.
Example 2.21
If the temperature of a dry air parcel at p
d
= 800 hPa is T = 270 K, then
p
=
287.8 K from (2.95).
A parameter related to potential temperature is potential virtual temperature
(
v
), which is found by converting all the moisture in a parcel to dry air, then
bringing the parcel to a pressure of 1000 hPa and determining its temperature.
v
is
the potential temperature of a sample of moist air as if it were dry and at the same
density and pressure as the moist air. It is derived by substituting
a
= R
T
v
p
a
into
(2.88), then integrating from T
0
=
v
to T and p
a.0
= 1000 hPa to p
a
. The result is
v
= T (1 +0.608q
v
)
1000 hPa
p
a
= T
v
1000 hPa
p
a
(2.96)
51
280 290 300 310 320 330
700
750
800
850
900
950
1000
Potential temperature (K)
P
r
e
s
s
u
r
e

(
h
P
a
)
3:30 a.m.
3:30 p.m.
Figure 2.12 Observed vertical pro-
les of potential temperature at
3:30 a.m. and 3:30 p.m. on August
27, 1987 at Riverside, California.
Potential temperatures correspond
to actual temperatures shown in
Fig. 2.11(a) and (b), respectively.
The change in entropy of an air parcel per unit mass (J kg
1
K
1
) due to
energy transfer between the parcel and its environment is dS = dQT. During adi-
abatic expansion and compression, dQ = 0 and no change in entropy occurs. Since
dQ = 0 along surfaces of constant potential virtual temperature, dS = 0 along such
surfaces as well, and the surfaces are called isentropic. Figure 2.13 illustrates that
potential virtual temperatures increase monotonically with height in the Northern-
Hemisphere troposphere, causing isentropic surfaces to slant toward the North
Pole.
A parameter related to potential virtual temperature is virtual potential tem-
perature, which is found by bringing a moist parcel to a pressure of 1000 hPa,
then converting all moisture in the parcel to dry air and determining the parcels
temperature. It is the virtual temperature of air that has been brought adiabatically
to 1000 hPa. It is obtained by applying the virtual temperature correction from
(2.38) to the potential temperature of moist air. The result is
p.v
=
p.m
(1 +0.608q
v
) = T
v
1000 hPa
p
a
(10.251q
v
)
(2.97)
The difference between potential virtual temperature and virtual potential temper-
ature is relatively small (Brutsaert 1991).
The Exner function, which is used in future chapters, is dened as c
p.d
P, where
P =
p
a
1000 hPa
(2.98)
52
Latitude
N. Pole
Equator
q
v,4
q
v,3
q
v,2
q
v,1
Isentropic surfaces
A
l
t
i
t
u
d
e
Decreasing q
v
I
n
c
r
e
a
s
i
n
g

q
v
Figure 2.13 Isentropic surfaces (surfaces of constant
potential virtual temperature) between the Equator
and the North Pole. Sea-level temperature and, there-
fore,
v
decrease from the Equator to the pole. Since
the free troposphere is stably stratied with respect to
unsaturated air,
v
increases with altitude, and lines of
constant
v
slant toward the poles in the vertical.
Substituting P into (2.95) and (2.96) gives
T
v
=
v
P (2.99)
2.6.2 Stability criteria for unsaturated air
The concepts derived above from the rst law of thermodynamics are useful for
analyzing atmospheric stability. The atmosphere is stable (stably stratied) when
a parcel of air, displaced vertically, decelerates and returns to its original position.
The atmosphere is unstable when a displaced parcel accelerates in the direction in
which it is displaced. The atmosphere is neutral when a parcel does not accelerate
or decelerate after being displaced.
When the atmosphere is stable near the surface, pollution builds up, since air
parcels cannot accelerate out of the stable layer to disperse the pollution. Stability
also inhibits clouds of vertical development from forming. When the atmosphere
is unstable, emitted pollutants accelerate vertically, decreasing their concentration
near the surface. Clouds of vertical development can form in unstable air.
2.6.2.1 Determining stability from the dry adiabatic lapse rate
When the air is unsaturated with water vapor (i.e., the relative humidity is less
than 100 percent), stability can be determined by comparing the environmental
lapse rate with the dry adiabatic lapse rate (I
d.m
from (2.90)). The environmental
lapse rate, I
e
= Tz, is the negative change of actual temperature with altitude.
When temperature decreases with increasing altitude, I
e
> 0. In terms of I
e
and
I
d.m
, the stability criteria for unsaturated air are
I
e
> I
d.m
unsaturated unstable
= I
d.m
unsaturated neutral
- I
d.m
unsaturated stable
(2.100)
53
0.8
1
1.2
1.4
1.6
1.8
2
2.2
0 5 10 15 20 25 30
A
l
t
i
t
u
d
e

(
k
m
)
Temperature (C)
Stable
Unstable
Figure 2.14 Demonstration of stability and
instability in unsaturated air. When a parcel is
displaced vertically, it rises and cools adiabati-
cally (along the dashed line). If the environmen-
tal temperature prole is stable (right thick line),
the rising parcel is cooler than the environment,
decelerates, then sinks to its original position. If
the environmental temperature prole is unsta-
ble (left thick line), the rising parcel is warmer
than the air around it and continues to rise. The
parcel stops rising only when it encounters air
with the same temperature as itself. This occurs
when the parcel reaches a layer with a newenvi-
ronmental temperature prole.
Figure 2.14 demonstrates how stability can be determined graphically.
Example 2.22
If the observed change in temperature with altitude is Tz =15 K km
1
and the air contains no water vapor, what is the stability class of the
atmosphere?
SOLUTION
The environmental lapse rate in this case is I
e
= +15 K km
1
, which is greater
than I
d.m
= +9.8 K km
1
; thus, the atmosphere is unstable with respect to
unsaturated air.
Equation (2.100) indicates that temperature can increase or decrease with
increasing altitude in stable air but must decrease with increasing altitude in
unstable air. A temperature inversion is an increase in temperature with increas-
ing altitude. An inversion occurs only in stable air, but the presence of sta-
ble air does not necessarily mean that an inversion is present, as illustrated in
Example 2.23.
54
Example 2.23
When temperature decreases slightly with increasing altitude (e.g., I
e
=
+2.0 Kkm
1
), the atmosphere is stable but an inversion is not present. An
inversion occurs when temperature increases with altitude, as demonstrated
by the line marked stable in Fig. 2.14.
Stability is enhanced by any process that warms air at higher altitudes relative to
air at lower altitudes. At night, surface air becomes stable because the ground cools
radiatively. During the day, stability is enhanced over land near the sea when cool
marine air travels inland and displaces warm land-air vertically (creating warm air
over cold air). Stability also increases when warm air blows over a cold surface or
when air in a high-pressure system sinks, compresses, and warms on top of cool
marine air below. Instability occurs when the land heats rapidly during the day or
when a cold wind blows over a warm surface.
2.6.2.2 Determining stability from potential virtual temperature
Stability can also be determined frompotential virtual temperature. Differentiating
(2.96) gives
d
v
= dT
v
1000
p
a
+ T
v
1000
p
a
1000
p
2
a
dp
a
=

v
T
v
dT
v
v
p
a
dp
a
(2.101)
Taking the partial derivative of (2.101) with respect to height and substituting
p
a
z =
a
g and the virtual temperature lapse rate, I
v
= T
v
z, into the
result give
v
z
=

v
T
v
T
v
z

v
p
a
p
a
z
=
v
T
v
I
v
+
R
c
p.d
v
p
a
a
g (2.102)
Substituting T
v
= p
a
a
R
and I
d
= gc
p.d
from (2.89) into (2.102) results in
v
z
=
v
T
v
I
v
+

v
g
T
v
c
p.d
=

v
T
v
(I
d
I
v
) (2.103)
Equation (2.100) indicates that the air is stable when I
d
> I
v
, suggesting that
v
z > 0 also indicates that the air is stable. Thus, the stability criteria in terms
of potential virtual temperature are
v
z
- 0 unsaturated unstable
= 0 unsaturated neutral
> 0 unsaturated stable
(2.104)
Figure 2.15 demonstrates howstability can be determined froma graph of potential
virtual temperature versus altitude.
55
0.8
1
1.2
1.4
1.6
1.8
2
2.2
10 15 20 25 30 35 40
A
l
t
i
t
u
d
e

(
k
m
)
Potential virtual temperature (C)
Stable
Unstable
Figure 2.15 Demonstration of stability and
instability from potential virtual temperature.
When an unsaturated parcel is displaced verti-
cally, it rises and cools adiabatically (along the
dashed line). If the ambient potential-virtual-
temperature prole slopes positively, a ris-
ing parcel is always cooler than the environ-
ment, the atmosphere is stable, and the parcel
sinks back to its original position. If the ambi-
ent potential-virtual-temperature prole slopes
negatively, a rising parcel is always warmer
than the environment, the atmosphere is unsta-
ble, and the parcel continues rising.
The advantage of deriving stability from
v
instead of frompotential temperature
of dry air is that the latter parameter does not account for water vapor whereas
the former does. The advantage of deriving stability from
v
instead of from the
potential temperature of moist air or from virtual potential temperature is that,
although the latter parameters include water vapor more accurately than does
v
, their differentiation requires several additional terms of little consequence not
required in the differentiation of
v
.
Example 2.24
Given I
v
= +7 K km
1
, p
a
= 925 hPa, and T
v
= 290 K, nd
v
z.
SOLUTION
From (2.96),
v
= 296.5 K. Thus, from (2.103),
v
z = 2.86 K km
1
, and the
atmosphere is stable with respect to unsaturated air.
2.6.2.3 Determining stability from the BruntV ais al a frequency
Another way to write equation (2.103) is
ln
v
z
=
1
T
v
(I
d
I
v
) (2.105)
56
2.7 Summary
Multiplying by gravitational acceleration, g, gives
N
2
bv
= g
ln
v
z
=
g
T
v
(I
d
I
v
) (2.106)
where N
bv
is the BruntV ais al a frequency (or buoyancy frequency).
The BruntV ais al a frequency measures the static stability of the atmosphere. If
v
increases with increasing height (I
d
> I
v
), then N
2
bv
> 0, and the atmosphere
is stable. In such a case, buoyancy acts as a restoring force, causing a perturbed
parcel of air to oscillate about its initial altitude with a period
bv
= 2N
bv
. During
the oscillation, kinetic energy is exchanged with potential energy. The oscillations
arising from buoyancy restoration are gravity waves, discussed in Chapter 4.
If
v
is constant with increasing altitude (N
2
bv
= 0), the atmosphere is neutral,
and displacements occur without resistance from a restoring buoyancy force. If
v
decreases with increasing altitude (N
2
bv
- 0), the atmosphere is unstable, and a
parcels displacement increases exponentially with time. In sum, the stability criteria
from the BruntV ais al a frequency are
N
2
bv
- 0 unsaturated unstable
= 0 unsaturated neutral
>0 unsaturated stable
(2.107)
Example 2.25
Given I
v
=+6.5 K km
1
and T
v
=288 K, estimate the BruntV ais al a frequency
and the period of oscillation of a perturbed parcel of air.
SOLUTION
Since I
d
= +9.8 K km
1
and g = 9.81 ms
2
, we have N
bv
= 0.0106 s
1
from
(2.106), and
bv
= 593 s. The atmosphere is statically stable with respect to
unsaturated air in this case.
2.7 SUMMARY
In this chapter, the structure and composition of the atmosphere were discussed.
The bottom 100 km of the atmosphere consists of four primary regions of
temperature variation: the troposphere, stratosphere, mesosphere, and thermo-
sphere. The troposphere consists of the boundary layer and the free troposphere.
Three important variables describing the atmosphere, temperature, pressure, and
density, are related by the equation of state. Other variables discussed include
the relative humidity and potential virtual temperature. Some equations derived
included the hydrostatic equation, ClausiusClapeyron equation, and rst law of
thermodynamics. The equation of state and hydrostatic equation were combined
57
to give expressions for altitude as a function of air pressure and vice versa. The
ClausiusClapeyron equation was combined with empirical relationships for latent
heat to derive expressions for the saturation vapor pressure of water over liquid
and ice. The rst law of thermodynamics was used to derive atmospheric stability
criteria. In the next chapter, the continuity and thermodynamic energy equations
are discussed.
2.8 PROBLEMS
2.1 If T = 295 K at 1 mm above the ground and the conductive heat ux is H
c
=
250 W m
2
, estimate the temperature at the ground. Assume the air is dry.
2.2 If N
q
= 1.5 10
12
molec. cm
3
for O
3
gas, T = 285 K, and p
d
= 980 hPa, nd
the volume mixing ratio, dry-air mass mixing ratio, and partial pressure of
ozone.
2.3 If
q
= 1.3 ppmm for carbon monoxide gas, T = 285 K, and p
d
= 980 hPa,
nd the volume mixing ratio, number concentration, and partial pressure of
carbon monoxide.
2.4 If T = 268 K and p
d
= 700 hPa, respectively, nd p
v.s
in hPa, and nd the
corresponding mixing ratio of water vapor in percent, ppmm, and ppmv.
2.5 Find the mass density of moist air (
a
) if T = 283 K, f
r
= 78 percent, and
p
d
= 850 hPa.
2.6 Find the pressure exerted by moist air if T = 288 K, f
r
= 82 percent, and
p
d
= 925 hPa.
2.7 Find the virtual temperature when N
a
= 2.1 10
19
molec. cm
3
, T = 295 K,
and f
r
= 92 percent.
2.8 Find the partial pressure of water vapor if q
v
= 3 g kg
1
, T = 278 K, and
d
= 0.5 kg m
3
.
2.9 If T
v
= 281 K, p
v
= 3 hPa, and p
a
= 972 hPa, nd the air temperature.
2.10 If the total air pressure, temperature, and relative humidity are p
a
=945 hPa,
T = 276 K, and f
r
= 46 percent, nd
v
, m
a
, R
m
, T
v
, and
a
.
2.11 If dry-air pressure, temperature, and water-vapordry-air mass mixing ratio
are p
d
= 927 hPa, T = 281 K, and
v
= 0.005 kg kg
1
, nd f
r
, m
a
, R
m
, T
v
, and
a
.
2.12 If the total air pressure, water-vapor volume mixing ratio, and temperature
are p
a
= 966 hPa,
v
= 3000 ppmv, and T = 284 K, nd p
v
, m
a
, R
m
, T
v
, and
a
.
2.13 Find the altitude in a standard atmosphere that a pressure altimeter gives if
the pressure measured by the altimeter is p
a
= 770 hPa and the air is dry.
2.14 Estimate the scale height of the atmosphere (H) and resulting pressure at
z = 200 m altitude if the air is dry, the pressure at z = 100 m is p
d
=
990 hPa, and the average temperature between z = 100 m and z = 200 m is
T = 284 K.
2.15 If the air is dry, z = 10 km, p
d
= 250 hPa, and T = 218 K (base of the
tropopause), estimate the scale height at z = 10 km and pressure at z =
10.5 km.
58
2.8 Problems
2.16 Calculate the saturation vapor pressure of water over liquid and ice if
T
c
= 15

C. Find
v
and p
v
at this temperature if f
r
= 3 percent and
p
d
= 230 hPa.
2.17 Calculate the dew point if f
r
= 54 percent and T = 263 K.
2.18 Derive the expression for the water-vapor mass mixing ratio as a function of
dew point and pressure from (2.68). If T
D
= 284 K and p
d
= 1000 hPa, nd
v
.
2.19 If T
D
= 279 K, T = 281 K, and p
d
= 930 hPa, calculate f
r
, p
v
,
v
, and
v.s
.
2.20 (a) Estimate the diabatic energy (dQ) that needs to be added to or removed
from a parcel of air to increase its virtual temperature dT
v
= +2 K when
a
= 1.2 kg m
3
and when the pressure change in the parcel due to
adiabatic expansion is dp
a
= 10 hPa.
(b) If dQ is removed from part (a) and other conditions stay the same,
what is the parcel virtual-temperature change? What type of process is
this?
(c) If the parcel in part (a) does not rise or expand (dp
a
= 0), but dQ calcu-
lated from part (a) remains, what is the virtual-temperature change of the
parcel? What is the name of this type of process?
(d) If the parcel in part (a) does not change virtual temperature (dT
v
=0), but
dp
a
= 10 hPa, what is the new value of dQ? What is the name of this
process?
2.21 Calculate the potential virtual temperature of dry air when (a) p
a
= 900 hPa
and T = 280 K; (b) p
a
= 850 hPa and T = 278 K. Is the air parcel between
900 and 850 hPa stable, unstable, or neutral with respect to unsaturated
air?
2.22 If
p
= 303 K at p
d
= 825 hPa, nd the air temperature at this pressure.
2.23 Calculate the change in potential virtual temperature with altitude (
v
z)
when the ambient virtual-temperature lapse rate is I
v
= +6.2 K km
1
, the
air pressure is p
a
= 875 hPa, and T
v
= 283 K. Is this air stable, unstable, or
neutral with respect to unsaturated air?
2.24 If the air is dry and the potential virtual temperature increases at the rate
v
z = 1 K km
1
, calculate the ambient virtual-temperature lapse rate
when p
a
= 925 hPa and T = 288 K. Is this air stable, unstable, or neutral
with respect to unsaturated air?
2.25 If the potential virtual temperature increases at the rate
v
z = 2 K km
1
in dry air at an altitude where p
a
= 945 hPa and T = 287 K, estimate T at
100 m above this altitude.
2.26 Would liquid water and/or ice particles grow when (a) p
v
= 1 hPa, T =
30 K; (b) p
v
= 1.2 hPa, T = 20 K; or (c) p
v
= 1 hPa, T = 16 K? Use
Fig. 2.8(b).
2.27 Does potential virtual temperature at sea level increase, decrease, or stay con-
stant (on average) between the Equator and North Pole? Why? Does potential
virtual temperature increase, decrease, or stay constant with altitude if I
v
=
+6.5 K km
1
? Why?
2.28 What might Figs. 2.3(a) and (b) look like under a low-pressure system, if all
other conditions were the same?
59
2.9 COMPUTER PROGRAMMING PRACTICE
2.29 Write a computer script to calculate air pressure (p
a
) as a function of alti-
tude from (2.41). Assume p
a
= 1013 hPa and T = 288 K at the surface, the
temperature decreases at the rate of 6.5 K km
1
, and the air is dry. Use the
programto estimate the pressure fromz =0 to 10 kmin increments of 100 m.
Calculate the density
a
with the equation of state at the base of each layer
before each pressure calculation for the next layer. Plot the results.
2.30 Write a computer script to calculate p
a
as a function of altitude from (2.44).
a
= 1013 hPa at the surface and the air is dry. Use
the program to estimate the pressure from z = 0 to 10 km in increments of
100 m. Plot the results.
a
as a function of altitude from (2.48).
a
= 1013 hPa at the surface, the air is dry, and the
temperature decreases from the surface at 6.5 K km
1
. Use the program to
estimate the pressure fromz =0 to 10 kmin increments of 100 m, calculating
the scale height for each layer. Plot the results.
2.32 Write a computer script to calculate the saturation vapor pressure of water
over liquid and ice from (2.61) and (2.64), respectively. Use the program to
estimate p
v.s
between 50

C and +50

C and p
v.I
between 50

C and 0

C,
in increments of 1

C. Plot the results.
v.s
, p
v.I
, and T
D
versus altitude. Assume
T is 298 K at z = 0 km and decreases 6.5 K km
1
. Assume also that f
r
= 90
percent at all altitudes. Use the program to estimate parameters from z = 0
to 10 km in increments of 100 m. Plot the results.
60
3
The continuity and thermodynamic
energy equations
T
he continuity equations for air, individual gases, and aerosol particles, and the
thermodynamic energy equation are fundamental equations in atmospheric
models. Continuity equations are used to simulate changes in concentration or
mixing ratio of a variable over time and take account of transport, external sources,
and external sinks of the variable. The thermodynamic energy equation is used to
predict changes in temperature with time and takes account of transport, external
sources, and external sinks of energy. In this chapter, scalars, vectors, gradient
operators, local derivatives, and total derivatives are dened, and the continuity
and thermodynamic energy equations are derived.
3.1 DEFINITIONS
In this section, denitions relating to wind speed and direction and differentiation
are provided. The denitions will be used in subsequent sections to derive time-
dependent continuity equations.
3.1.1 Wind velocity
Scalars are variables, such as temperature, air pressure, and air mass, that have
magnitude but not direction. Vectors are variables, such as velocity, that have
magnitude and direction.
Winds are described by three parameters velocity, the scalar components of
velocity, and speed. Velocity is a vector that quanties the rate at which the position
of a body changes over time. Total and horizontal wind velocity vectors are dened
in Cartesian (rectangular) horizontal coordinates as
v = iu +jv +kw v
h
= iu +jv (3.1)
respectively, where i, j, and k are Cartesian-coordinate westeast, southnorth, and
vertical unit vectors, respectively, and
u =
dx
dt
v =
dy
dt
w =
dz
dt
(3.2)
are scalar components of velocity (scalar velocities) (m s
1
). Scalar velocities have
magnitude only. When applied in (3.1), positive u, v, and w correspond to winds
61
The continuity and thermodynamic energy equations
0
0
40
20
0
20
20
40
-40
-20
0 0
0
-60
60
20
80 60 40 20 0 20 40 60 80
100
80
60
40
20
0
Latitude (deg)
A
l
t
i
t
u
d
e

(
k
m
)
(a)
30
20
40
20
0
0
20
40
60
40
80
-60
-40
-20
0
0
80 60 40 20 0 20 40 60 80
100
80
60
40
20
0
Latitude (deg)
A
l
t
i
t
u
d
e

(
k
m
)
(b)
Figure 3.1 Zonally averaged westeast scalar velocities (ms
1
) for (a) January
and (b) July. Data for the plots were compiled by Fleming et al. (1988).
moving from west to east, south to north, and lower to higher elevation, respec-
tively. The vertical scalar velocity in (3.2) is written in the altitude vertical coordi-
nate system. In this coordinate system(coordinate), tops and bottoms of horizontal
layers are dened by surfaces of constant altitude.
The magnitude of the wind is its speed. The total and horizontal wind speeds
are dened as
|v| =
u
2
+v
2
+w
2
|v
h
| =
u
2
+v
2
(3.3)
respectively.
Wind direction is generally named for where a wind originates from. A westerly
wind, southwesterly wind, sea breeze, and mountain breeze originate fromthe west,
the southwest, the sea, and a mountain, respectively. A positive scalar velocity u
with no southnorth component is a westerly wind. A positive scalar velocity v
with no westeast component is a southerly wind.
Air velocities vary in space and time. Figures 3.1(a) and (b) show global-scale
latitudealtitude contour plots of zonally averaged westeast scalar velocities for
January and July, respectively. The gures indicate that westeast winds in the
upper troposphere almost always originate from the west. The two peaks near
10 km in each gure correspond to subtropical jet streams. Near the surface at the
Equator and poles, winds originate from the east but are weak. Near the surface at
midlatitudes (30
60
) in both hemispheres, winds originate from the west. In the

stratosphere, westerly wind speeds increase with height in the winter hemisphere
(Northern Hemisphere in January; Southern Hemisphere in July), forming polar
night jets near 60 km. Easterly wind speeds increase with increasing altitude in the
summer hemisphere. Winds above the surface are driven by pressure gradients, and
pressure gradients are driven by temperature gradients. Thus, strong winds aloft
indicate strong temperature and pressure gradients.
62
3.1 Definitions
3.1.2 Time and spatial rates of change
The time rate of change of a variable, such as concentration, momentum, or tem-
perature, can be determined at a xed location or in the frame of reference of the
variable as it moves. Suppose a plume, carrying a gas with number concentration
N= N(t. x[t]) (molec. cm
3
), travels with the wind from xed point A in the west
to xed point B in the east. The time rate of change of Nanywhere along the plumes
trajectory is the total derivative, dNdt. The total derivative can be expanded with
the chain rule in Cartesian coordinates as
dN
dt
=
N
t
dt
dt
+
N
x
dx
dt
=
N
t
+u
N
x
(3.4)
where Nt is the time rate of change of concentration at xed point A (local
derivative), and u Nx is the time rate of change of concentration in the plume
that results from a westeast scalar velocity transporting the plume.
The total derivative of a variable is nonzero when processes other than transport
affect the variable. In the case of gases, external processes include chemistry and
gas-to-particle conversion. If dNdt =0, the concentration of a gas does not change
as it travels with the wind.
The local derivative of a variable is the difference between the total derivative
and the rate of change of the variable due to transport. Thus, the local derivative is
affected by external processes plus transport. If Nt = 0, the rate of production
(loss) of a variable due to external processes equals the rate of loss (production) of
the variable due to transport of a spatial gradient of the variable [u( Nx)].
Example 3.1
Suppose the time rate of change of concentration of a gas along the path of a
hot-air balloon traveling with the wind from east to west at u = 10 m s
1
is dNdt = 10
8
molec. cm
3
s
1
. If the westeast gradient in concentration is
Nx = 10
10
molec. cm
3
km
1
(concentration increases from west to east),
determine the time rate of change of concentration at a xed point A, which
the balloon passes over.
SOLUTION
Since u Nx = 10
8
molec. cm
3
s
1
, (3.4) predicts ( Nt)
A
2 10
8
molec. cm
3
s
1
. Thus, transport from the east (u Nx) accounts for one-half
of the production rate of Nat point A, and transformations along the trajectory
(dNdt) account for the other half.
A Lagrangian frame of reference is a frame of reference that moves relative to a
xed coordinate system. An Eulerian frame of reference is a frame of reference in
a xed coordinate system. The left side of (3.4) is written in terms of a Lagrangian
63
frame of reference. The right side is written in terms of an Eulerian frame of refer-
ence. Generalizing (3.4) to three dimensions gives
dN
dt
=
N
t
+u
N
x
+v
N
y
+w
N
z
(3.5)
3.1.3 Gradient operator
Agradient operator (also called a directional derivative, nabla operator, or del oper-
ator) is a vector operator of partial derivatives. The gradient operator in Cartesian-
altitude coordinates is
= i

x
+j

y
+k

z
(3.6)
The dot product of the velocity vector with the gradient operator is a scalar
operator,
v
= (iu +jv +kw)
i

x
+j

y
+k

z
= u

x
+v

y
+w

z
(3.7)
where i
i = 1, j
j = 1, and k
k = 1. Cross terms are zero (i
j = 0, i
k = 0, and
j
k = 0), since the unit vectors are orthogonal. The dot product of two vectors is
a scalar and symmetric (e.g., a
v = v
a). The dot product of a gradient operator

with a vector is a scalar operator but not symmetric (
v = v
). Instead,
v =
i

x
+j

y
+k

z
(iu +jv +kw) =

u
x
+
v
y
+
w
z
(3.8)
which is a scalar divergence. When concentration is multiplied by a divergence, the
result is the scalar
N(
v) = N
u
x
+ N
v
y
+ N
w
z
(3.9)
The gradient of a scalar, such as concentration, is a vector. For example,
N=
i

x
+j

y
+k

z
N= i
N
x
+j
N
y
+k
N
z
(3.10)
Applying the dot product of velocity with the gradient operator to N gives the
scalar
(v
)N=
u

x
+v

y
+w

z
N= u
N
x
+v
N
y
+w
N
z
(3.11)
Substituting this result into the total-derivative equation (3.5) yields
dN
dt
=
N
t
+(v
)N (3.12)
64
3.2 Continuity equations
z
u
1
N
1
u
2
N
2
y
x
Figure 3.2 Example of mass con-
servation. The number of molecules
entering minus the number of
molecules leaving the box equals the
number of molecules accumulating
in the box.
Generalizing (3.12) for any variable gives the total derivative in Cartesian-altitude
coordinates as
d
dt
=

t
+u

x
+v

y
+w

z
=

t
+v
(3.13)
3.2 CONTINUITY EQUATIONS
When air circulates in an enclosed volume, and no chemical or physical processes
affect it, the mass of the air, summed throughout the volume, is conserved. In
an atmospheric model divided into many grid cells (grid boxes), the mass of air
entering one cell minus the mass leaving the cell equals the nal mass minus the
initial mass in the cell. The same is true for other atmospheric variables, such as
gas concentrations or energy, when only transport affects these variables.
Figure 3.2 shows a grid cell with dimensions Lx, Ly, Lz (m). The westeast
scalar velocities entering and leaving the cell are u
1
and u
2
(m s
1
), respectively.
Gas concentrations at the west and east boundaries of the cell are N
1
and N
2
(molec. cm
3
), respectively. Mass uxes of gas into the cell and out of the cell are
u
1
N
1
and u
2
N
2
(m molec. cm
3
s
1
), respectively.
From the information given, the numbers of molecules entering, leaving, and
accumulating in the box during time period Lt are u
1
N
1
LyLzLt, u
2
N
2
LyLzLt,
and
LNLxLyLz = u
1
N
1
LyLzLt u
2
N
2
LyLzLt (3.14)
respectively. Dividing both sides of (3.14) by Lt and by the box volume (LxLyLz)
gives
LN
Lt
=
u
2
N
2
u
1
N
1
Lx
(3.15)
65
As Lx 0 and Lt 0, this equation approaches
N
t
=
(uN)
x
(3.16)
which is the continuity equation for a gas affected by velocity in one direction. This
equation expands to three dimensions in Cartesian-altitude coordinates as
N
t
=
(uN)
x

(vN)
y

(wN)
z
=
(vN) (3.17)
where vN= iuN+jvN+kwN. A similar equation can be written for air density.
Equation (3.17) states that the time rate of change of N at a xed location equals
the negative of the local spatial gradient of the ux of N. Equation (3.17) is a
ux divergence form of the continuity equation so called because
(vN) is a
divergence of concentration.
Substituting
(vN) = N(
v) +(v
)N (3.18)
into (3.17) and writing a similar equation for air density give the continuity equa-
tions for gas number concentration and total air mass density as
N
t
= N(
v) (v
)N (3.19)
a
t
=
a
(
v) (v
)
a
(3.20)
respectively. Substituting
(v
)N=
dN
dt

N
t
(3.21)
from (3.12) into (3.19) and (3.20) gives velocity divergence forms of the continuity
equations as
dN
dt
= N(
v) (3.22)
d
a
dt
=
a
(
v) (3.23)
where
v is the divergence of velocity. The equations are also advective forms of

the continuity equation in that ddt contains the advection term, v
. The equa-
tions state that the change of a scalar variable over time in a moving parcel equals
the scalar variable multiplied by the negative local spatial gradient of velocity.
The gas number concentration N(molecules per cubic centimeter of air) is related
to the moist-air mass mixing ratio, q (kilograms per kilogram of moist air), of a
66
3.2 Continuity equations
species with molecular weight m (g mol
1
) by
N=
A
a
q
m
(3.24)
where A is Avogadros number (molec. mol
1
). Substituting (3.24) into (3.19) and
expanding give
q
a
t
+
a
(
v) +(v
)
a
+
a
q
t
=
a
(v
)q (3.25)
Substituting the continuity equation for air from (3.20) into (3.25) gives the gas
continuity equation in units of the moist-air mass mixing ratio as
q
t
= (v
)q (3.26)
Equations (3.22) and (3.23) assume that air is compressible, meaning that total
volume of a parcel of air changes over time. Ocean water is considered incompress-
ible, meaning that the total volume of a parcel of ocean water does not change over
time. Thus,
u
x
+
v
y
+
w
z
= 0 (3.27)
which is the continuity equation for an incompressible uid. If (3.27) is not satis-
ed, a net divergence out of or convergence into a uid volume occurs, causing the
volume to expand or contract, respectively. Equation (3.27) states that an incom-
pressible uid is nondivergent. The equation can also be written as
v = 0. Sub-
stituting water density (
w
) for air density and substituting
v = 0 into (3.23)
give
d
w
dt
= 0 (3.28)
which states that the density of an incompressible uid is constant along the motion
of the uid. At a xed point in the uid, the density may change. Substituting water
density for air density and
v = 0 into (3.20) give
w
t
= (v
)
w
(3.29)
which states that the change in water density at a xed point in an incompressible
uid is the negative product of velocity and the spatial gradient of density. In sum,
the density of an incompressible uid, such as liquid water, can vary spatially, but
the total volume of such a uid is constant over time. Auid in which density varies
spatially throughout the uid is inhomogeneous. Ocean water is inhomogeneous
and incompressible. A uid in which density is always constant throughout a vol-
ume (
w
t = 0) is homogeneous. Pure liquid water is a relatively homogeneous,
incompressible uid. Air is an inhomogeneous and compressible uid.
67
3.3 EXPANDED CONTINUITY EQUATIONS
Equation (3.17) gave the continuity equation without molecular diffusion or exter-
nal source and sink terms. A more complete form of the continuity equation for a
gas is
N
t
=
(vN) + D
2
N+
N
e.t
n=1
R
n
(3.30)
(e.g., Reynolds et al. 1973), where D is the molecular diffusion coefcient of the
gas (cm
2
s
1
), N
e.t
is the number of external processes (e.g., chemistry, emission,
etc.) affecting the gas, and R
n
is the time rate of change of trace-gas concentration
due to the nth external process affecting the gas (molec. cm
3
s
1
). Molecular
diffusion is the movement of molecules due to their kinetic energy. As molecules
move, they collide with other molecules and are redirected along arbitrary paths.
A molecular diffusion coefcient quanties the rate of molecular diffusion, and is
dened mathematically in Section 16.2.
The squared gradient in the molecular diffusion term of (3.30) expands to
2
N= (
) N =
i

x
+j

y
+k

z
i

x
+j

y
+k

z
N
=

2
N
x
2
+

2
N
y
2
+

2
N
z
2
(3.31)
Substituting (3.17) and (3.31) into (3.30) gives the continuity equation for a gas as
N
t
+
(uN)
x
+
(vN)
y
+
(wN)
z
= D
2
N
x
2
+

2
N
y
2
+

2
N
z
2
+
N
e.t
n=1
R
n
(3.32)
3.3.1 Time and grid volume averaging
The spatial domain in a model is divided into grid cells of nite size. Time is also
divided into time steps of nite size for advancing species concentrations, veloc-
ities, and other variables. Real atmospheric motions generally occur over spatial
scales much smaller than the resolution of model grid cells and over temporal scales
smaller than the resolution of model time steps. For example, a typical mesoscale
model might have horizontal resolution 5 km 5 km, vertical resolution 50 m,
and time resolution 5 s. A global-scale model might have horizontal resolution
400 km 400 km, vertical resolution 200 m, and time resolution 300 s. Fluctua-
tions in atmospheric motions due to eddies occur on smaller scales in both cases.
Eddies (Section 4.2.6; Section 8.4), for example, range in diameter from a couple
of millimeters to hundreds of meters and on time scales of seconds to hours.
Whereas many models do not resolve eddies, some do. These are discussed
in Section 8.4. To account for subgrid-scale disturbances in those models that
do not resolve eddies, a process called Reynolds averaging, named after Osborne
Reynolds, is used. Models that treat turbulence using Reynolds averaging are called
Reynolds-averaged models.
68
3.3 Expanded continuity equations
During Reynolds averaging, each variable in (3.32) and in other model equations
is divided into an average and perturbation component. Such a division is referred
to as Reynolds decomposition. In the case of gases, gas number concentration is
decomposed as
N=

N+ N
(3.33)
where N is the actual (precise or instantaneous) concentration,

N is the average
concentration, and N
is the instantaneous perturbation concentration. A precise

concentration occurs at a given instant and location within a grid cell. An average
concentration is obtained by integrating and averaging over a model time step and
grid-cell volume. Thus,
N=
1
hLxLyLz
t+h
t
x+Lx
x
y+Ly
y
z+Lz
z
N dz
dy
dx
dt (3.34)
(e.g., Pielke 1984), where h is the time step, and Lx, Ly, Lz are space increments,
shown in Fig. 3.2. The average concentrations are averages over one grid cell and
time step and differ for each grid cell and time step. Perturbation concentrations are
distributed on both sides of the average, so that the spatial and temporal average
of all perturbations is zero (

N
= 0), which is the Reynolds assumption.

Scalar and vector velocities can be decomposed in a similar manner. Thus, for
example,
u = u +u
v = v +v
w = w+w
(3.35)
where u, v, and w are the time- and volume-averaged scalar velocities, and u
, v
,
and w
are perturbation scalar velocities, and

v = v +v
(3.36)
where v = i u +j v +k wis a time- and volume-averaged velocity and v
= iu
+jv
+
kw
is a perturbation velocity. Advection is the mean horizontal velocity. Thus, u

and v are components of advection. Figure 3.3 shows an example of precise, mean,
and perturbation scalar velocities and trace-gas concentrations.
Unsteady ow occurs when v varies with time, but not necessarily randomly, at
a given location. Steady ow occurs when v is independent of time. Turbulent ow
is unpredictable ow in which v varies randomly with time at a location. Thus,
turbulent ow is unsteady, but unsteady ow is not necessarily turbulent. Laminar
ow is nonturbulent ow in which v may vary, but not randomly, with time at a
given location. In laminar ow, uid particles travel along well-dened streamlines
and uid layers ow independent of each other. Laminar ow can be steady or
unsteady. Nearly all ows in the atmosphere are turbulent.
Subgrid-scale effects are estimated by substituting decomposed variables into an
equation, then taking a time average and grid volume average of resulting terms.
Substituting (3.33) and (3.35) into the species continuity equation from (3.32) and
69
N'
u'
u
N
u
_
N
_
Figure 3.3 Precise, mean, and perturbation components of scalar velocity
and gas concentration. The precise scalar velocity is denoted by u, the precise
gas concentration is denoted by N, time- and volume-averaged values are
denoted by an overbar, and perturbation components are denoted by a
prime. Each point on the horizontal axis is a perturbation at a given time
and location within a grid cell.
averaging terms over space and time (Reynolds averaging) give
(

N+ N
)
t
( u +u
)(

N+ N
)
x
( v +v
)(

N+ N
)
y
( w+w
)(

N+ N
)
z
= D
2
(

N+ N
)
x
2
2
(

N+ N
)
y
2
2
(

N+ N
)
z
2
+
N
e.t
n=1
R
n
(3.37)
Since (

N+ N
)t = (

N+ N
)t,

N+ N
N+

N
N=

N, and

N
= 0, the
rst term in (3.37) simplies to
(

N+ N
)
t
=
(

N+

N
)
t
=

N
t
(3.38)
Since u

N= 0, uN
= 0, and u

N= u

N, the second term simplies to
( u +u
)(

N+ N
)
x
=
( u

N+ uN
+u

N+u
)
x
=
( u

N+u
)
x
(3.39)
The product u
(m molec. cm
3
s
1
) represents the westeast transport of N
due
to subgrid-scale eddies. It is a kinematic turbulent ux in that its units are those
of concentration ux (kg [molec. cm
3
] m
2
s
1
) divided by air density (kg m
3
).
When a variable or a ux is divided by the air density, it becomes a kinematic
variable or ux. The partial derivative (u
)x (molec. cm
3
s
1
) is a turbulent
ux divergence term.
70
Example 3.2
Suppose two gas concentrations (N
1
= 8 and N
2
= 4) and scalar velocities
(u
1
= 3 and u
2
= 1) are measured at different locations within a grid cell at
a given time. Estimate

N, N
1
, N
2
, u, u
1
, u
2
, u
, u

N, and uN. Ignore units.
SOLUTION
N = (N
1
+ N
2
)2 = 6 u = (u
1
+u
2
)2 = 1
N
1
= N
1

N = 2 u
1
= u
1
u = 2
N
2
= N
2

N = 2 u
2
= u
2
u = 2
u
= (u
1
N
1
+u
2
N
2
)2 = 4 u

N = 6
uN = u

N+u
= (u
1
N
1
+u
2
N
2
)2 = 10
Substituting (3.38), (3.39), and similar terms for other directions into (3.37)
gives

N
t
+
( u

N)
x
+
( v

N)
y
+
( w

N)
z
+
u
x
+
v
y
+
w
z
= D
2

N
x
2
+

2

N
y
2
+

2

N
z
2
+
N
e.t
n=1
R
n
(3.40)
For motions larger than the molecular scale, the molecular diffusion terms in (3.40)
are much smaller than are the turbulent ux divergence terms and can be removed.
Thus, (3.40) simplies to

N
t
+
( u

N)
x
+
( v

N)
y
+
( w

N)
z
+
u
x
+
v
y
+
w
z
=
N
e.t
n=1
R
n
(3.41)
which compresses to the continuity equation for a gas,

N
t
+
( v

N) +
(v
) =
N
e.t
n=1
R
n
(3.42)
An analogous equation for air density results in the continuity equation for air,

a
t
+
(v
a
) +
(v
a
) = 0 (3.43)
In (3.43), the external source and sink terms for air molecules are neglected because
they are small in comparison with the other terms.
71
Equation (3.42) can be rederived in terms of moist-air mass mixing ratio. Adding
source and sink terms to (3.26) yields
q
t
+v
q =
N
e.t
n=1
R
n
(3.44)
where R
n
is now in kilograms per kilogram of moist air per second. Multiplying
the continuity equation for air from (3.20) by q, multiplying (3.44) by
a
, adding
the results, and compressing give
(
a
q)
t
+
(
a
vq) =
a
N
e.t
n=1
R
n
(3.45)
The moist-air mass mixing ratio, velocity, and density can be decomposed with
q = q +q
, v = v +v
, and
a
=
a
+
a
, respectively. Density perturbations in the
atmosphere are relatively small; thus,
a

a
, and
a

a
. Substituting decom-
posed variable values into all but the R term in (3.45) gives
[
a
( q +q
)]
t
+

a
( v +v
)( q +q
=
a
N
e.t
n=1
R
n
(3.46)
Taking the time and grid volume average of this equation, eliminating zero-value
terms and removing unnecessary overbars results in

a
q
t
+( v
) q
+ q

a
t
+
(v
a
)
(
a
v
) =
a
N
e.t
n=1
R
n
(3.47)
Equation (3.47) can be simplied by rst noting that, when
a

a
, (3.43) becomes

a
t
+
(v
a
) = 0 (3.48)
Substituting this expression into (3.47) and dividing through by
a
give
q
t
+( v
) q +
1

a
(
a
v
) =
N
e.t
n=1
R
n
(3.49)
which is the gas continuity equation in units of the moist-air mass mixing ratio.
In (3.49), u
, v
, and w
(m kg kg
1
s
1
) are kinematic turbulent uxes of
mixing ratio. Whereas models calculate spatially and temporally averaged values
(e.g.,

N, q,
a
, u), kinematic turbulent uxes (e.g., u
) are parameterized. Some

parameterizations are discussed in Section 8.4. Here, a simple overviewof K-theory
is given.
With K-theory (gradient transport theory), kinematic turbulent uxes are
replaced with the product of a constant and the gradient of the mean value of a uc-
tuating variable (Calder 1949; Pasquill 1962; Monin and Yaglom 1971; Reynolds
et al. 1973; Stull 1988). This is convenient, because models predict mean quantities.
72
Kinematic turbulent uxes of gas concentration, for example, are parameterized
with
u
= K
h.xx

N
x
v
= K
h.yy

N
y
w
= K
h.zz

N
z
(3.50)
where K
h,xx
, K
h,yy
, and K
h,zz
(e.g., cm
2
s
1
) are eddy diffusion coefcients in the
x-, y-, and z-directions, respectively. The subscript h indicates that the eddy dif-
fusion coefcient for energy (eddy thermal diffusivity) is used. The eddy diffusion
coefcient for energy is used because the turbulent transport of a gas is similar to
that of energy. When turbulent transport of velocity is simulated, an eddy diffusion
coefcient for momentum (eddy viscosity) term is used. Eddy diffusion coefcients
for energy and momentum differ, but not by much. Eddy diffusion coefcients
represent an average diffusion coefcient for eddies of all sizes smaller than the
grid cell. Eddy diffusion coefcients are also called eddy transfer, eddy exchange,
turbulent transfer, and gradient transfer coefcients.
Eddy diffusion coefcients are parameterizations of subgrid scale transport of
energy and momentum. In the vertical, such transport is caused by mechanical
shear (mechanical turbulence) and/or buoyancy (thermal turbulence). Horizontal
wind shear creates eddies that increase in size when the wind ows over rough sur-
faces. Buoyancy creates instability, causing shear-induced eddies to become wider
and taller. Vertical motions in eddies transfer surface air upward and air aloft
downward. Eddies also exchange air horizontally.
Substituting (3.50) into (3.41) gives

N
t
+
( u

N)
x
+
( v

N)
y
+
( w

N)
z
=

x
K
h.xx

N
x
+

y
K
h.yy

N
y
+

z
K
h.zz

N
z
+
N
e.t
n=1
R
n
(3.51)
Compressing (3.51) and removing overbars for simplicity give the continuity equa-
tion for an individual gas in number concentration units and Cartesian-altitude
coordinates as
N
t
+
(vN) = (
K
h
) N+
N
e.t
n=1
R
n
(3.52)
where
K
h
=
K
h.xx
0 0
0 K
h.yy
0
0 0 K
h.zz
(3.53)
73
is the eddy diffusion tensor for energy. The analogous continuity equation for an
individual gas in moist-air mass mixing ratio units is
q
t
+(v
)q =
1
a
(
a
K
h
)q +
N
e.t
n=1
R
n
(3.54)
The units of R
n
differ in the two cases.
3.3.2 Continuity equation for air
External sources and sinks (R
n
) are relatively small and can be ignored in the
continuity equation for air. For most modeling applications, the turbulent ux
divergence term in the equation can also be ignored because
a

a
. After remov-
ing overbars for convenience and making the above modications, the continuity
equation for air in Cartesian-altitude coordinates reduces from (3.43) to
a
t
+
(v
a
) = 0 (3.55)
3.3.3 Gas continuity equation
The continuity equations for trace gases and particles include several external
source and sink terms. Gases enter the atmosphere from surface and elevated
sources by emission. They are removed onto water, soil, foliage, roads, buildings,
cars, and other surfaces by dry deposition. In many cases, gases are swept out of
the atmosphere by falling raindrops during washout. Gases react chemically with
each other and are dissociated by solar radiation during photochemistry. Some
gases aggregate to form new particles during homogeneous nucleation or aggre-
gate on existing particle surfaces during heterogeneous nucleation. Once a surface
has nucleated, gas molecules may diffuse to and condense as a liquid or deposit as
a solid on the surface. Liquid material may also evaporate or solid material may
sublimate to the gas phase. A gas may also dissolve in liquid water on the surface
of a particle. Dissolved gases may evaporate. Finally, a gas may react chemically
on the surface of a particle during heterogeneous chemistry.
A form of the continuity equation for a gas q that accounts for the processes
discussed above is
N
q
t
+
(vN
q
) = (
K
h
)N
q
+ R
emisg
+ R
depg
+ R
washg
+ R
chemg
+ R
nucg
+ R
c/eg
+ R
dp/sg
+ R
ds/eg
+ R
hrg
(3.56)
74
where
R
emisg
= rate of surface or elevated emission
R
depg
= rate of dry deposition to the ground
R
washg
= rate of washout to the ground or from one altitude to another
R
chemg
= rate of photochemical production or loss
R
nucg
= rate of gas loss due to homogeneous or heterogeneous nucleation
R
c/eg
= rate of gas loss (production) due to condensation (evaporation)
R
dp/sg
= rate of gas loss (production) due to depositional growth (sublimation)
R
ds/eg
= rate of gas loss (production) due to dissolutional growth (evaporation)
R
hrg
= rate of gas loss (production) due to heterogeneous reactions
All rates are expressed in units of concentration per unit time (e.g., molec. cm
3
s
1
).
3.3.4 Particle continuity equation
The continuity equation for particles is divided into two subequations. One is
for particle number concentration, and the other is for particle volume compo-
nent concentration. Particles contain anywhere from one to hundreds of com-
ponents. The volume of each component varies over time due to physical and
chemical processes. If the total volume of one particle in a size bin i is denoted
by
i
(cm
3
/particle), the volume of component q within that particle is
q.i
.
Thus,
q.i
gives information about a component in a single particle of a given
size. A variable giving information about that component summed over all par-
ticles of the same size, is more relevant. Such a parameter is volume concentra-
tion (cubic centimeters of the component per cubic centimeter of air), dened
as
v
q.i
= n
i
q.i
(3.57)
where n
i
is the number concentration of particles of size i (particles per cubic
centimeter of air). If two of the three variables in (3.57) are predicted numerically,
the third can be found from the equation. Typically, volume concentration and
number concentration are predicted numerically. They are found from separate
continuity equations, because different external sources and sinks affect the number
and volume concentrations.
The continuity equation for the number concentration of particles of size i is
n
i
t
+
(vn
i
) = (
K
h
) n
i
+ R
emisn
+ R
depn
+ R
sedn
+ R
washn
+ R
nucn
+ R
coagn
(3.58)
75
where
R
emisn
R
depn
= rate of particle dry deposition to the surface
R
sedn
= rate of sedimentation to the surface or between layers
R
washn
= rate of washout to the surface or from one altitude down to another
R
nucn
= rate of production of new particles due to homogeneous nucleation
R
coagn
= rate of coagulation of number concentration
All rates are in units of particles cm
3
s
1
. Sources and sinks that affect particle
number concentration include emission, dry deposition, sedimentation, washout,
homogeneous nucleation, and coagulation. Sedimentation occurs when particles
fall through the atmosphere due to their mass. Sedimentation by gases is negligible
because gas molecules have extremely small masses. Particle dry deposition occurs
when particles diffuse to or otherwise impact a surface by any transport process.
Particle washout occurs when rain sweeps particles in its path to lower altitudes or
the surface. Homogeneous nucleation is a source of new particles. Heterogeneous
nucleation does not produce newparticles but allows growth to proceed on existing
particles. Coagulation occurs when two particles collide and stick to form a single,
larger particle.
The continuity equation for the volume concentration of component q in parti-
cles of size i is
v
q.i
t
+
(vv
q.i
) = (
K
h
) v
q.i
+ R
emisv
+ R
depv
+ R
sedv
+ R
washv
+ R
nucv
+ R
coagv
+ R
c/ev
+ R
dp/sv
+ R
ds/ev
+ R
eqv
+ R
aqv
+ R
hrv
(3.59)
where
R
emisv
R
depv
= rate of dry deposition to the surface
R
sedv
= rate of sedimentation to the surface or from one altitude to another
R
washv
= rate of washout to the surface or from one altitude to another
R
nucv
= rate of change due to homogeneous or heterogeneous nucleation
R
coagv
= rate of change due to coagulation
R
c/ev
= rate of change due to condensational growth (evaporation)
R
dp/sv
= rate of change due to depositional growth (sublimation)
R
ds/ev
= rate of change due to dissolutional growth (evaporation)
R
eqv
= rate of change due to reversible chemical equilibrium reactions
R
aqv
= rate of change due to irreversible aqueous chemical reactions
R
hrv
= rate of change due to heterogeneous reactions on particle surfaces
Rates in this equation have units of cubic centimeters of component q per cubic
centimeter of air per second. Some processes, such as homogeneous nucleation
and coagulation, affect number and volume concentrations. Others, such as
76
heterogeneous nucleation, condensation, deposition, dissolution, heterogeneous
reaction, chemical equilibrium and aqueous chemistry affect volume concentra-
tion but not number concentration.
3.3.5 Continuity equation for gas, liquid, and solid water
Water in the atmosphere appears as a gas, liquid, or solid. In a model, the total
water content is estimated as
q
T
= q
v
+
N
B
i =1
(q
L.i
+q
I.i
) (3.60)
where N
B
is the number of particle size categories (bins), q
v
is the specic humidity
of water vapor (kilograms per kilogram of moist air), q
L.i
is the moist-air mass
mixing ratio of liquid water in a size bin, and q
I.i
is the moist-air mass mixing
ratio of ice in a size bin. Mass mixing ratios are determined from the continuity
equations for water vapor, liquid, and ice,
q
v
t
+(v
)q
v
=
1
a
(
a
K
h
)q
v
+ R
emisV
+ R
depV
+ R
chemV
+ R
c/eV
+ R
dp/sV
(3.61)
q
L.i
t
+(v
)q
L.i
=
1
a
(
a
K
h
)q
L.i
+ R
emisL
+ R
depL
+ R
sedL
+ R
coagL
+ R
c/eL
+ R
f/mL
(3.62)
q
I.i
t
+(v
)q
I.i
=
1
a
(
a
K
h
)q
I.i
+ R
depI
+ R
sedI
+ R
coagI
+ R
f/mI
+ R
dp/sI
(3.63)
where
R
emis
R
dep
= rate of dry deposition to the surface
R
sed
= rate of sedimentation to the surface or from one altitude to another
R
chem
= rate of photochemical production or loss
R
coag
= rate of liquid or ice production or loss in a size bin due to coagulation
R
c/e
= rate of change due to condensational growth (evaporation)
R
dp/s
= rate of change due to depositional growth (sublimation)
R
f/m
= rate of change due to freezing (melting)
and the units of R are kilograms per kilogram of moist air per second.
Many meteorological models simulate liquid water and ice as bulk parame-
ters. In such cases, liquid water and ice are not separated into size categories,
and their number concentrations are not tracked. Instead, only the moist-air mass
mixing ratios of total liquid water and ice are predicted. Since particles are not size-
resolved in a bulk parameterization, some processes, such as coagulation, cannot
be treated adequately. When liquid and ice content are treated as bulk parameters,
77
q
T
= q
v
+q
L
+q
I
, where the subscript i has been dropped because a bulk param-
eterization has no size resolution.
3.4 THERMODYNAMIC ENERGY EQUATION
Air temperature is affected by energy transfer and work. Energy transfer processes
include conduction, mechanical turbulence, thermal turbulence, advection, and
radiation, all introduced in Section 2.2. Energy is released to the air during conden-
sation of water vapor, deposition of water vapor, freezing of liquid water, exother-
mic chemical reactions, and radioactive decay. Energy is removed fromthe air upon
melting of ice, sublimation of ice, and evaporation of liquid water. Energy exchange
may also occur upon the change of state of substances other than water. Because
the quantities of nonwater substances changing state are relatively small, resulting
energy exchanges are small. Energy, like air density and species concentrations, is
conserved in a system.
An equation describing energy changes in the atmosphere can be derived by
combining the rst law of thermodynamics with the continuity equation for air.
The rst lawof thermodynamics as expressed in (2.82) was dQ c
p.d
dT
v
a
dp
a
.
Differentiating this equation with respect to time, substituting
a
= 1
a
, and rear-
ranging give the thermodynamic energy equation as
dT
v
dt

1
c
p.d
dQ
dt
+
1
c
p.d
a
dp
a
dt
(3.64)
If the thermodynamic energy equation is written in terms of potential
virtual temperature, the last term in (3.64) can be eliminated. Differentiating
v
= T
v
(1000p
a
)
with respect to time give

d
v
dt
=
dT
v
dt
1000
p
a
+ T
v
1000
p
a
1000
p
2
a
dp
a
dt
=

v
T
v
dT
v
dt

v
p
a
dp
a
dt
(3.65)
Substituting (3.65), = R
c
p.d
, and p
a
=
a
R
T
v
into (3.64), and expanding the
total derivative with (3.13) give the thermodynamic energy equation in terms of
potential virtual temperature as
d
v
dt
=

v
t
+(v
)
v

v
c
p.d
T
v
dQ
dt
(3.66)
Multiplying all terms in (3.66) by c
p.d
a
, multiplying all terms in the continuity
equation for air from (3.20) by c
p.d
v
, adding the two equations, and compressing
give
(c
p.d
v
)
t
+
(vc
p.d
v
)
a
v
T
v
dQ
dt
(3.67)
78
3.4 Thermodynamic energy equation
Substituting the energy density (J m
3
), dened as E = c
p.d
v
, into (3.67) gives
the continuity equation for energy,
E
t
+
(vE)
a
v
T
v
dQ
dt
(3.68)
This equation is similar to the continuity equations for air mass density or gas
number concentration. It states that the time rate of change of energy in a box
equals the energy ux in minus the energy ux out plus (minus) external sources
(sinks). Replacing N with E in Fig. 3.2 yields a diagram of energy uxes into and
out of a hypothetical grid cell.
In a model, subgrid eddies affect energy transport. To account for such eddies,
variables in (3.67) can be decomposed as v = v +v
,
a
=
a
+
a
and
v
=

v
+
v
.
Since density perturbations are small (
a

a
), density is approximated as
a

a
.
Substituting velocity, density, and potential virtual temperature decompositions
into (3.67), setting the approximation to an equal sign for simplicity, and taking
the time- and grid-volume average of the result yield
c
p.d
[
a
(
v
+
v
)]
t
+c
p.d
[
a
( v
v
+ v
v
+v

v
+v
v
)] =
a
v
T
v
dQ
dt
(3.69)
Eliminating zero-value time and spatial derivatives and unnecessary overbars
results in
(
a
v
)
t
+
(
a
v
v
) +
(
a
v
v
) =

a
c
p.d
v
T
v
dQ
dt
(3.70)
which expands to

a
v
t
+( v

a
t
+
(v
a
)
(
a
v
v
) =

a
c
p.d
v
T
v
dQ
dt
(3.71)
Substituting the continuity equation for air from (3.48) into (3.71) and dividing by

a
give the thermodynamic energy equation as
v
t
+( v
v
+
1

a
(
a
v
v
) =

v
c
p.d
T
v
dQ
dt
(3.72)
The kinematic turbulent sensible-heat uxes (v
v
) can be parameterized with
u
v
= K
h.xx
v
x
v
v
= K
h.yy
v
y
w
v
= K
h.zz
v
z
(3.73)
where the eddy diffusion coefcients for energy are the same as those used in the
continuity equation for a trace species. Substituting v
v
= K
h
v
into (3.72) and
eliminating overbars for simplicity give
v
t
+(v
)
v
=
1
a
(
a
K
h
)
v
+

v
c
p.d
T
v
dQ
dt
(3.74)
79
The diabatic heating rate consists of the terms
dQ
dt
=
N
e.h
n=1
dQ
n
dt
=
dQ
c/e
dt
+
dQ
f/m
dt
+
dQ
dp/s
dt
+
dQ
solar
dt
+
dQ
ir
dt
(3.75)
where N
e.h
is the number of diabatic energy sources and sinks. All Qs are in joules
per kilogram. dQ
c/e
dt is the rate of energy release (absorption) due to condensation
(evaporation), dQ
f/m
dt is the rate of energy release (absorption) due to freezing
(melting), dQ
dp/s
dt is the rate of energy release (absorption) due to deposition
(sublimation), dQ
solar
dt is the rate of solar heating, and dQ
ir
dt is the rate of net
infrared heating (cooling). Substituting (3.75) into (3.74) gives the thermodynamic
energy equation as
v
t
+(v
)
v
=
1
a
(
a
K
h
)
v
+

v
c
p.d
T
v
N
e.h
n=1
dQ
n
dt
(3.76)
3.5 SUMMARY
In this chapter, local and total derivatives were dened, and the continuity and ther-
modynamic energy equations were derived. Continuity equations included those
for air, trace gases, aerosol number concentration, and aerosol volume concentra-
tion. These equations treat subgrid eddy motions with kinematic turbulent ux
terms, which are generally parameterized. A common type of parameterization
is a K-theory parameterization. Equations described in this chapter are necessary
for simulating the transport and transformations of total air, gases, aerosol par-
ticles, and energy. An equation used for predicting wind speed and direction, the
momentum equation, is discussed next.
3.6 PROBLEMS
3.1 Expand the total derivative of the u-scalar velocity (i.e., substitute u for N
in (3.5)) when the air ow is (a) steady, (b) unsteady.
3.2 Explain why (3.17) differs from (3.26).
3.3 What is the purpose of Reynolds averaging?
3.4 If u = 5 m s
1
and v = +5 m s
1
, write out the horizontal velocity vector,
determine the horizontal wind speed, and name the wind.
3.5 Assume that a grid cell has dimension Lx = 5 km, Ly = 4 km, and Lz =
0.1 km and that the west, east, south, north, and lower scalar velocities are
u
1
= +3, u
2
= +4, v
3
= 3, v
4
= +2, and w
5
= +0.2 m s
1
. If the atmosphere
is incompressible, what is w at the top of the cell?
3.6 (a) A grid cell has dimensions Lx = 5 km, Ly = 4 km, and Lz = 0.1 km.
Assume the gas concentration and scalar velocity at the west boundary
of the cell are N
1
= 110
11
molec. cm
3
and u
1
= +7 m s
1
, respectively,
and those at the east boundary of the cell are N
2
= 5 10
11
molec. cm
3
and u
2
= +8 m s
1
, respectively. (i) Assuming external sources and sinks
80
3.7 Computer programming practice
and eddy diffusion are absent, estimate N at the cell center after 60 s if
the initial N is an average of the two boundary N-values and boundary
parameters remain constant. (ii) Calculate the time after the start at which
N at the cell center becomes zero.
(b) Assume that the gas concentration and scalar velocity at the south bound-
ary of the grid cell in part (a) are N
3
= 1 10
11
molec. cm
3
and v
3
=
2 m s
1
, respectively, and those at the north boundary are N
4
= 7
10
11
molec. cm
3
and v
4
= +1 m s
1
, respectively. Calculate (i) N at the
cell center after 60 s and (ii) the time after the start at which N at the
center becomes zero. Assume uxes operate in four directions, and the
initial N at the center of the cell is the average of all four grid-boundary
N-values.
(c) Convert the gas number concentrations frompart (a) (N
1
and N
2
) to moist-
air mass mixing ratios, assuming that the gas is ozone, T
v
= 298 K, and
p
a
= 1013 hPa.
(d) Re-solve parts (a) (i) and (a) (ii) with (3.26) using moist-air mass mixing
ratios instead of number concentration units. Assume that the westeast
velocity for this question is an average of the grid-cell boundary veloc-
ities. Convert the mass mixing ratio from the 60-s case back to number
concentration units. Howdoes the result compare with that found in part
(a) (i)? If it differs, why does it differ?
3.7 Agrid cell has dimensions Lx =5 km, Ly =4 km, and Lz =0.1 km. Assume
that the potential virtual temperature, pressure, and scalar velocity at the
west boundary of the grid cell are
v.1
= 302 K, p
a.1
= 1004 hPa, and u
1
= +7
m s
1
, respectively, and those at the east boundary of the grid cell are
v.2
=
299 K, p
a.2
= 1008 hPa, and u
2
= +8 m s
1
, respectively.
(a) Calculate the virtual temperature and air density at the west and east
boundaries of the grid cell.
(b) Calculate the energy density E at each boundary.
(c) Assuming diabatic energy sources and sinks and eddy diffusion are
absent, calculate the potential virtual temperature at the center of the
grid cell after 10 s.
3.8 Assume that grid-cell size, boundary conditions, and Nare initially the same
as in Problem 3.6(a). Write a computer script to calculate the nal N at the
grid-cell center after a time step h. After each time step, set the east-boundary
gas concentration (N
2
) equal to the nal concentration at the center of the
cell. Set h = 3 s, and run the program for a simulation period of one hour.
Plot the grid center concentration versus time.
3.9 Assume that grid-cell size, boundary conditions, and initial
v
are the same
as in Problem 3.7. Write a computer script to calculate the nal
v
at the
cell center after a time step h. After each step, set
v.2
equal to
v
at the cell
center. Set h = 3 s, and run the program for six hours. Plot
v
versus time at
the cell center.
81
4
The momentum equation in Cartesian
and spherical coordinates
T
he momentumequation (equation of motion) describes the movement of air. In
a model, it is used to predict wind velocity (speed and direction). In this chapter,
the momentum equation and terms within it are derived. These terms include local
acceleration, the Earths centripetal acceleration (apparent centrifugal force), the
Coriolis acceleration (apparent Coriolis force), the gravitational force, the pressure-
gradient force, the viscous force, and turbulent-ux divergence. The equation is
derived for Cartesian and spherical horizontal coordinate systems. Cartesian coor-
dinates are often used over microscale and mesoscale domains, where the Earths
curvature may be neglected. Spherical coordinates are used over global-, synoptic-,
and many mesoscale, domains, where curvature cannot be neglected. Equations for
the geostrophic wind, gradient wind, and surface wind are derived fromthe momen-
tum equation. Atmospheric waves are also discussed. Important waves include
acoustic, Lamb, gravity, inertia Lamb, inertia gravity, and Rossby waves.
4.1 HORIZONTAL COORDINATE SYSTEMS
Atmospheric modeling equations can be derived for a variety of horizontal coordi-
nate systems. In this section, some of these systems are briey discussed. Equations
for the conversion of variables from Cartesian to spherical horizontal coordinates
are then given.
4.1.1 Cartesian, spherical, and other coordinate systems
Many atmospheric models use Cartesian or spherical horizontal coordinates. Carte-
sian (rectangular) coordinates are used on the microscale and mesoscale to simulate
ow, for example, in street canyons, downwind of smokestacks, in cities, and in
clouds. Over short distances (-500 km), the Earths curvature is relatively small,
and Earths surface is often divided into rectangles for modeling. Over long dis-
tances, curvature prevents the accurate division of the Earths surface into a contigu-
ous set of rectangles. Nevertheless, over such distances, it is possible to envelop the
Earth with many rectangular meshes, each with a different origin and nite overall
length and width, and where each mesh partly overlaps other meshes. This is the
idea behind the universal transverse Mercator (UTM) coordinate system, which is
a type of Cartesian coordinate system. In the UTM system, separate, overlapping
meshes of rectangular grid cells are superimposed over the globe. UTM coordinate
82
4.1 Horizontal coordinate systems
i
l
k
r
R
e
cos j
e
R
e
R
e
j
j
j
Figure 4.1 Spherical coordi-
nate symbols. R
e
is the Earths
radius, is latitude,
e
is lon-
gitude, and i
, j
, and k
r
are
westeast, southnorth, and
vertical unit vectors, respec-
tively.
locations are mapped back to spherical coordinate locations with UTM-to-
spherical conversion equations (U.S. Department of the Army 1958).
For model simulations on small or large scales, the use of spherical coordinates
(Fig. 4.1) is more natural than the use of Cartesian coordinates. The spherical coor-
dinate system divides the Earth into longitudes (meridians), which are southnorth
lines extending from the South Pole to the North Pole, and latitudes (parallels),
which are westeast lines parallel to each other extending around the globe. The
Prime Meridian, which runs through Greenwich, United Kingdom, is dened to
have longitude 0
. Meridians extend westward to 180
(180W) longitude and

eastward to +180
(180E) longitude. The Equator is dened to have latitude 0
.
Parallels extend from90
(90S) latitude to +90
(90N) latitude. On the spherical-

coordinate grid, the westeast distance between meridians is the greatest at the
Equator and converges to zero at both poles. In fact, all meridians converge to
a single point at the poles. Thus, the poles are singularities. The presence of a
singularity at the poles presents a boundary-condition problem when the spheri-
cal coordinate system is used for global atmospheric or ocean (in the case of the
North Pole since no ocean exists over the South Pole) simulations. This problem
is addressed in Chapter 7. The main advantage of the spherical coordinate sys-
tem is that it takes into account the Earths curvature since the Earth is close to
spherical.
Some other horizontal map projections are Mercator, stereographic, and Lam-
bert conformal. A Mercator projection is one in which each rhumb line on a sphere
is represented as a straight line. A rhumb line is a curve on the surface of a sphere
that cuts all meridians at the same angle. Since the angle can be any angle, a rhumb
line can spiral to the poles. A stereographic projection is one in which points on a
sphere correspond exactly to points on an extended plane and in which the North
Pole on the sphere corresponds to innity on the plane. A Lambert conformal
83
Momentum equation in Cartesian and spherical coordinates
projection is one in which meridians are represented as straight lines converging
toward the nearer pole and parallels are represented as arc segments of concentric
circles. Although these three projections are used in cartography, they are used less
frequently in atmospheric modeling. Snyder (1987) presents equations for convert-
ing among these and other coordinate systems.
Spherical and Cartesian coordinate grids are regular grids. In a regular grid, grid
cells are aligned in a lattice or xed geometric pattern. Such grid cells do not need
to be rectangular. For example, spherical-coordinate grid cells are nonrectangular
but are distributed in a xed pattern. Another type of grid is an irregular grid. In
an irregular grid, grid cells are not aligned in a lattice or xed pattern and may
have irregular shape and size. Irregular grids are useful for modeling applications
in which real boundaries do not match up well with regular-grid boundaries. For
example, in ocean modeling, coastlines are uneven boundaries that do not match
up well with Cartesian or spherical coordinate boundaries. Irregular grids are also
useful for modeling the North and South Poles in a global model to avoid the sin-
gularity problem that arises with the spherical coordinate system. Finally, irregular
grids are useful for treating some regions in a model at high resolution and others
at lower resolution to save computer time.
Regular grids can also be applied to some regions at high resolution and others at
low resolution through the use of grid stretching and nesting. Grid stretching is the
gradual decrease then increase in westeast and/or southnorth grid spacing on a
spherical-coordinate grid to enable higher resolution in some locations. Nesting
is the placement of a ne-resolution grid within a coarse-resolution grid that
provides boundary conditions to the ne-resolution grid. Nesting is discussed in
Chapter 21.
Flows over irregular grids are generally solved with nite-element methods
(Section 6.5) or nite-volume methods (Section 6.6). Finite-difference methods
(Section 6.4) are challenging (but not impossible) to implement over irregular
grids. Flows over regular grids are generally solved with nite-difference meth-
ods although the nite-element and nite-volume methods are often used as well.
Celia and Gray (1992) describe nite-element formulations on an irregular grid.
Durran (1999) describes nite-element and nite-volume formulations on an irreg-
ular grid. In this text, the formulation of the equations of atmospheric dynamics is
limited to regular grids.
4.1.2 Conversion from Cartesian to spherical coordinates
Figure 4.1 shows the primary components of the spherical coordinate system on
a spherical Earth. Although the Earth is an oblate spheroid, slightly bulging at
the Equator, the difference between the equatorial and polar radii is small enough
(21 km) that the Earth can be considered to be a sphere for modeling purposes.
The spherical-coordinate unit vectors for the Earth are i
. j
and k
r
, which are
westeast, southnorth, and vertical unit vectors, respectively. Because the Earths
surface is curved, spherical-coordinate unit vectors have a different orientation at
84
4.1 Horizontal coordinate systems
each horizontal location on a sphere. On a Cartesian grid, i, j, and k are oriented
in the same direction everywhere on the grid.
In spherical coordinates, westeast and southnorth distances are measured in
terms of changes in longitude (
e
) and latitude (), respectively. The vertical coor-
dinate for now is the altitude (z) coordinate. It will be converted in Chapter 5 to
the pressure (p), sigmapressure (p), and sigmaaltitude (z) coordinates.
Conversions between increments of distance in Cartesian coordinates and incre-
ments of longitude or latitude in spherical coordinates, along the surface of Earth,
are obtained from the equation for arc length around a circle. In the westeast and
southnorth directions, these conversions are
dx = (R
e
cos )d
e
dy = R
e
d (4.1)
respectively, where R
e
6371 km is the radius of the Earth, d
e
is a westeast
longitude increment (radians), d is a southnorth latitude increment (radians),
and R
e
cos is the distance from the Earths axis of rotation to the surface of the
Earth at latitude , as shown in Fig. 4.1. Since cos is maximum at the Equator
(where = 0) and minimum at the poles, dx decreases from Equator to pole when
d
e
is constant.
Example 4.1
If a grid cell has dimensions d
e
= 5
and d = 5
, centered at = 30
N
latitude, nd dx and dy at the grid cell latitudinal center.
SOLUTION
First, d
e
= d = 5
o
180
o
= 0.0873 radians. Substituting these values into
(4.1) gives dx = (6371) (0.866) (0.0873) = 482 km and dy = (6371) (0.0873) =
556 km.
The local velocity vector and local horizontal velocity vector in spherical coor-
dinates are
v = i
u +j
v +k
r
w v
h
= i
u +j
v (4.2)
respectively. When spherical coordinates are used, horizontal scalar velocities can
be redened by substituting (4.1) into (3.2), giving
u =
dx
dt
= R
e
cos
d
e
dt
v =
dy
dt
= R
e
d
dt
w =
dz
dt
(4.3)
The third term in (4.3) is the altitude-coordinate vertical scalar velocity, which is
the same on a spherical horizontal grid as on a Cartesian horizontal grid.
85
e
R
e
cos j
R
e
cos j
W
E
l
e
i
l
i
l
i
l
+i
l
i
l
x
j
R
e
R
e
N
S
-k
r
i
l
cos j
i
l
j
j
i
l
sin j
j
(a) (b)
Figure 4.2 (a) Polar and (b) equatorial views of the Earth, show-
ing unit vectors used to determine i
e
. Adapted from Holton
(1992).
The gradient operator in spherical-altitude coordinates is
= i
1
R
e
cos
e
+j
1
R
e
+k
r
z
(4.4)
which is found by substituting (4.1) into (3.6) and replacing Cartesian- with
spherical-coordinate unit vectors. In spherical coordinates, the dot product of the
gradient operator with a scalar can be written from (4.4) with no more than three
terms. However, the dot product of the gradient operator with a vector requires
more than three terms because unit vectors change orientation at different locations
on a sphere. For example, expanding
v in spherical coordinates gives
v =
1
R
e
cos
e
+j
1
R
e
+k
r
(i
u +j
v +k
r
w)
=
1
R
e
cos
u
e
+i
u
1
R
e
cos
i
e
+i
v
1
R
e
cos
j
e
+i
w
1
R
e
cos
k
r
1
R
e
v
+j
u
1
R
e
i
+j
v
1
R
e
j
+j
w
1
R
e
k
r
w
z
+k
r
u
i
z
+k
r
v
j
z
+k
r
w
k
r
z
(4.5)
where some terms were eliminated because i

j
= 0, i

k
r
= 0, and j

k
r
= 0.
Partial derivatives of the unit vectors in (4.5) can be derived graphically. From
Fig. 4.2(a) and the equation for arc length around a circle, we have
|Li
| = |i
| L
e
= L
e
(4.6)
where |Li
| is the magnitude of the change in the westeast unit vector per unit
change in longitude, L
e
. Figure 4.2(b) indicates that, when L
e
is small,
Li
= j
|Li
| sin k
r
|Li
| cos (4.7)
86
4.2 Newtons second law of motion
Substituting (4.6) into (4.7), dividing by L
e
, and letting Li
0 and L
e
0
give
i
L
e
sin k
r
L
e
cos
L
e
j
sin k
r
cos (4.8)
Similar derivations for other derivatives yield
i
e
= j
sin k
r
cos
i
= 0
i
z
= 0
j
e
= i
sin
j
= k
r
j
z
= 0
k
r
e
= i
cos
k
r
= j
k
r
z
= 0
(4.9)
Substituting these expressions into (4.5) gives
v =
1
R
e
cos
u
e
+
1
R
e
cos
(v cos ) +
1
R
2
e
wR
2
e
(4.10)
which simplies to
v =
1
R
e
cos
u
e
+
1
R
e
cos
(v cos ) +
w
z
(4.11)
when R
e
is held constant. Since the incremental distance z above the Earths surface
is much smaller (-60 km) for most modeling applications than the radius of Earth
(6371 km), the assumption of a constant R
e
for use in (4.11) gives only a small
error.
4.2 NEWTONS SECOND LAW OF MOTION
The momentumequation is derived fromNewtons second lawof motion, F = Ma,
where F is force (N), Mis mass (kg), and a is acceleration (ms
2
). Newtons second
law states that the acceleration of a body due to a force is proportional to the
force, inversely proportional to the mass of the body, and in the direction of the
force. When applied to the atmosphere, the second law can be written in vector
form as
a
i
=
1
M
a
F (4.12)
where a
i
is the total or inertial acceleration, which is the rate of change of velocity
of a parcel of air in motion relative to a coordinate system xed in space (outside
the Earthatmosphere system), M
a
is the mass of the air parcel, and
F is the sum
of the force vectors acting on the parcel. A reference frame at rest or that moves
in a straight line at a constant velocity is an inertial reference frame. A reference
frame that either accelerates or rotates is a noninertial reference frame. Aspaceship
accelerating or a car at rest, at constant velocity, or accelerating on a rotating. Earth
87
k
r
sin j
j
R
e
R
e
j
j
cos j
j
Figure 4.3 Components of the
Earths angular velocity vector.
is in a noninertial reference frame. An observer at a xed point in space is in an
inertial reference frame with respect to any body on a rotating Earth, even if the
body is at rest on the surface of Earth.
Inertial acceleration is derived by considering that, to an observer xed in space,
the absolute velocity (m s
1
) of a body in motion near the surface of the Earth is
v
A
= v +R
e
(4.13)
where v is the local velocity, dened in (4.2), of the body relative to the Earths
surface, is the angular velocity vector for Earth, R
e
is the radius vector for the
Earth, and R
e
is the rate of change in position of the body due to the Earths
rotation. The Earths angular velocity vector (rad s
1
) and radius vector (m) are
dened as
= j
O cos +k
r
O sin R
e
= k
r
R
e
(4.14)
where O = 2 rad86 164 s = 7.292 10
5
rad s
1
is the magnitude of the angu-
lar velocity, and 86 164 is the number of seconds that the Earth takes to make
one revolution around its axis (23 h 56 m 4 s). The angular velocity vector acts
perpendicular to the equatorial plane of the Earth, as shown in Fig. 4.3. It does not
have a westeast component.
Inertial acceleration is dened mathematically as
a
i
=
dv
A
dt
+v
A
(4.15)
Substituting (4.13) into (4.15) and noting that is independent of time give
a
i
=
dv
dt
+
dR
e
dt
+v +(R
e
) (4.16)
88
The total derivative of R
e
is
dR
e
dt
= R
e
dk
r
dt
= i
u +j
v v (4.17)
where dk
r
dt is derived shortly in (4.28). Substituting (4.17) into (4.16) yields
a
i
=
dv
dt
+2v +(R
e
) = a
l
+a
c
+a
r
(4.18)
where
a
l
=
dv
dt
a
c
= 2v a
r
= (R
e
) (4.19)
are the local, Coriolis, and Earths centripetal accelerations, respectively. Local
acceleration is the rate of change of velocity of a parcel of air in motion relative to
a coordinate system xed on Earth, Coriolis acceleration is the rate of change of
velocity of a parcel due to the rotation of a spherical Earth underneath the parcel,
and the Earths centripetal acceleration is the inward-directed rate of change of
velocity of a parcel due to its motion around the Earths axis.
When Reynolds decomposition is applied to the precise local acceleration term
in (4.19), the term becomes a
l
= a
l
+a
l
, where a
l
is a mean local acceleration and
a
l
is a perturbation component, called a turbulent-ux divergence of momentum.
This term accounts for perturbations to the mean ow of wind, such as those
caused by mechanical shear (mechanical turbulence), thermal buoyancy (thermal
turbulence), and atmospheric waves. For now, the precise local acceleration term
is retained. It will be decomposed later in this section.
Whereas the centripetal and Coriolis effects are viewed as accelerations from an
inertial frame of reference, they are viewed as apparent forces from a noninertial
frame of reference. An apparent (or inertial) force is a ctitious force that appears to
exist when an observation is made in a noninertial frame of reference. For example,
when a car rounds a curve, a passenger within, who is in a noninertial frame of
reference, appears to be pulled outward by a local apparent centrifugal force, which
is equal and opposite to local centripetal acceleration multiplied by mass. On the
other hand, an observer in an inertial frame of reference sees the passenger and car
accelerating inward as the car rounds the curve. Similarly, as the Earth rotates, an
observer in a noninertial frame of reference, such as on the Earths surface, views
the Earth and atmosphere being pushed away from the Earths axis of rotation
by an apparent centrifugal force. On the other hand, an observer in an inertial
frame of reference, such as in space, views the Earth and atmosphere accelerating
inward.
The Coriolis effect can also be viewed from different reference frames. In a
noninertial frame of reference, moving bodies appear to feel the Coriolis force
pushing themto the right in the Northern Hemisphere and to the left in the Southern
Hemisphere. In an inertial frame of reference, such as in space, rotation of the
89
Earth underneath a moving body makes the body appear to accelerate toward the
right in the Northern Hemisphere or left in the Southern Hemisphere. In sum, the
centripetal and Coriolis effects can be treated as either accelerations or apparent
forces, depending on the frame of reference considered.
The terms on the right side of (4.12) are real forces. Real forces that affect
local acceleration of a parcel of air include the force of gravity (true gravitational
force), the force arising from spatial pressure gradients (pressure-gradient force),
and the force arising from air molecules exchanging momentum with each other
(viscous force). Substituting inertial acceleration terms from (4.18) into (4.12) and
expanding the right side give
a
l
+a
c
+a
r
=
1
M
a
(F
g
+F
p
+F
v
) (4.20)
where F
g
represents true gravitational force, F
p
represents the pressure gradient
force, and F
v
represents the viscous force. Atmospheric models usually require
expressions for local acceleration; thus, the momentum equation is written most
conveniently in a reference frame xed on the surface of the Earth rather than
xed outside the Earthatmosphere system. In such a case, only local acceleration
is treated as an acceleration. The Coriolis acceleration is treated as a Coriolis force
per unit mass (a
c
= F
c
M
a
), and the Earths centripetal acceleration is treated as
an apparent centrifugal (negative centripetal) force per unit mass (a
r
= F
r
M
a
).
Combining these terms with (4.20) gives the momentum equation from a reference
frame xed on Earths surface as
a
l
=
1
M
a
(F
r
F
c
+F
g
+F
p
+F
v
) (4.21)
In the following subsections, terms in (4.21) are discussed.
4.2.1 Local acceleration
The local acceleration, or the total derivative of velocity, expands to
a
l
=
dv
dt
=
v
t
+(v
) v (4.22)
This equation states that the local acceleration along the motion of a parcel equals
the local acceleration at a xed point plus changes in local acceleration due to
uxes of velocity gradients.
In Cartesian-altitude coordinates, the left side of (4.22) expands to
dv
dt
=
d(iu +jv +kw)
dt
= i
du
dt
+j
dv
dt
+k
dw
dt
(4.23)
90
and the right side expands to
v
t
+(v
) v =

t
+u

x
+v

y
+w

z
(iu +jv +kw) (4.24)

= i
u
t
+u
u
x
+v
u
y
+w
u
z
+j
v
t
+u
v
x
+v
v
y
+w
v
z
+ k
w
t
+u
w
x
+v
w
y
+w
w
z
In spherical-altitude coordinates, the left side of (4.22) expands, with the chain
rule, to
dv
dt
=
d(i
u +j
v +k
r
w)
dt
=
du
dt
+u
di
dt
dv
dt
+v
dj
dt
k
r
dw
dt
+w
dk
r
dt
(4.25)
Time derivatives of the unit vectors are needed to complete this equation. Substi-
tuting (4.1) into the total derivative in Cartesian-altitude coordinates from (3.13)
gives the total derivative in spherical-altitude coordinates as
d
dt
=

t
+u
1
R
e
cos
e
+v
1
R
e
+w

z
(4.26)
Applying (4.26) to i
yields
di
dt
=
i
t
+u
1
R
e
cos
i
e
+v
1
R
e
i
+w
i
z
(4.27)
Since i
does not change in time at a given location, i
t = 0. Substituting
i
t = 0 and terms from (4.9) into (4.27) and into like expressions for dj
dt
and dk
r
dt gives
di
dt
= j
utan
R
e
k
r
u
R
e
(4.28)
dj
dt
= i
utan
R
e
k
r
v
R
e
dk
r
dt
= i
u
R
e
+j
v
R
e
Finally, substituting (4.28) into (4.25) results in
dv
dt
= i
du
dt

uv tan
R
e
+
uw
R
e
+j
dv
dt
+
u
2
tan
R
e
+
vw
R
e
+k
r
dw
dt

u
2
R
e
v
2
R
e
(4.29)
91
Example 4.2
If u = 20 m s
1
, : = 10 m s
1
, and n = 0.01 m s
1
, and if dudt scales as
u(Lxu), estimate the value of each term on the right side of (4.29) at =
45
N latitude assuming Lx = 500 km, Ly = 500 km, and Lz = 10 km for

large-scale motions.
SOLUTION
From the values given,
du
dt
8 10
4
ms
2
u: tan
R
e
3.1 10
5
ms
2
un
R
e
3.1 10
8
ms
2
d:
dt
2 10
4
ms
2
u
2
tan
R
e
6.3 10
5
ms
2
:n
R
e
1.6 10
8
ms
2
dn
dt
1 10
8
ms
2
u
2
R
e
6.3 10
5
ms
2
:
2
R
e
1.6 10
5
ms
2
unR
e
and :nR
e
are small for large- and small-scale motions. dndt is also
small for large-scale motions.
Example 4.2 shows that uwR
e
and vwR
e
are small for large-scale motions and
can be removed from(4.29). If these terms are removed, u
2
R
e
and v
2
R
e
must also
be removed from the vertical term to avoid a false addition of energy to the system.
Fortunately, these latter terms are small in comparison with the gravitational and
pressure-gradient forces per unit mass. Implementing these simplications in (4.29)
gives the local acceleration in spherical-altitude coordinates as
dv
dt
= i
du
dt

uv tan
R
e
+j
dv
dt
+
u
2
tan
R
e
+k
r
dw
dt
(4.30)
Expanding the total derivative in (4.30) gives the right side of (4.22) in spherical-
altitude coordinates as
v
t
+(v
) v = i
u
t
+
u
R
e
cos
u
e
+
v
R
e
u
+w
u
z

uv tan
R
e
+j
v
t
+
u
R
e
cos
v
e
+
v
R
e
v
+w
v
z
+
u
2
tan
R
e
+k
r
w
t
+
u
R
e
cos
w
e
+
v
R
e
w
+w
w
z
(4.31)
In the horizontal, local accelerations have magnitude on the order of 10
4
ms
2
.
These accelerations are less important than Coriolis accelerations or than the
pressure-gradient force per unit mass, but greater than accelerations due to the vis-
cous force, except adjacent to the ground. In the vertical, local accelerations over
large horizontal distances are on the order of 10
7
m s
2
and can be neglected,
92
N. Pole
Direction of the
earth's rotation
N. Pole
Actual path
Intended path
Intended path
Actual path
Figure 4.4 Example of Coriolis deections. The Coriolis
force deects moving bodies to the right in the Northern
Hemisphere and to the left in the Southern Hemisphere.
The deection is zero at the Equator. Deections in the
gure are exaggerated.
since gravity and pressure-gradient accelerations are a factor of 10
8
larger. Over
small horizontal distances (-3 km), local accelerations in the vertical are important
and cannot be ignored.
4.2.2 Coriolis force
The second term in the momentum equation is the Coriolis force. In a noninertial
frame of reference, the Coriolis force appears to push moving bodies to the right
in the Northern Hemisphere and to the left in the Southern Hemisphere. In the
Northern Hemisphere, it acts 90
to the right of the direction of motion, and in

the Southern Hemisphere, it acts 90
to the left of the direction of motion. The

Coriolis force is only apparent: no force really acts. Instead, the rotation of a
spherical Earth below a moving body makes the body accelerate to the right in
the Northern Hemisphere or left in the Southern Hemisphere when viewed from
an inertial frame of reference, such as from space. The acceleration is zero at
the Equator, maximum near the poles, and zero for bodies at rest. Moving bodies
include winds, ocean currents, airplanes, and baseballs. Figure 4.4 gives an example
of Coriolis deections.
In terms of an apparent force per unit mass, the Coriolis term in (4.19) expands
in spherical-altitude coordinates to
F
c
M
a
= 2v = 2O
k
r
0 cos sin
u v w
= i
2O(wcos v sin) +j
2Ousin k
r
2Oucos (4.32)
If only a zonal (westeast) wind is considered, (4.32) simplies to
F
c
M
a
= 2v = j
2Ousin k
r
2Oucos (4.33)
93
2u
k
r
2u cos j
j
j
2u sin j
j
R
e
R
e
Figure 4.5 Coriolis acceleration com-
ponents that result when the Coriolis
force acts on a west-to-east wind trav-
eling around the Earth, parallel to the
Equator, and into the page (denoted
by arrow tail). See Example 4.3 for a
discussion.
Example 4.3 uses (4.33) to show that moving bodies are deected to the right in
the Northern Hemisphere.
Example 4.3
The fact that the Coriolis effect appears to deect moving bodies to the right
in the Northern Hemisphere can be demonstrated graphically. Consider only
local acceleration and the Coriolis force per unit mass in (4.21). In such a
case, a
l
= F
c
M
a
. Substituting (4.33) and a
l
= dvdt from (4.19) into this
expression when a westeast wind is present gives
dv
dt
= j
2Ou sin +k
r
2Ou cos
Figure 4.5 shows the terms on the right side of this equation and the magnitude
and direction of the resulting acceleration. The gure shows that the Coriolis
effect acts perpendicular to a wind blowing from the west (+u), forcing the
wind toward the south and vertically. At =0
N, only the vertical component

of the Coriolis effect remains, and the wind is not turned horizontally. At
=90
N, only the horizontal component remains, and the wind is not turned
vertically.
Because vertical scalar velocities are much smaller than horizontal scalar veloc-
ities, i
2Owcos may be removed from (4.32). Because the vertical component of

the Coriolis force is smaller than other vertical components in the momentumequa-
tion (e.g., gravity and pressure-gradient terms), k
r
2Oucos may also be removed.
With these changes, the Coriolis force vector per unit mass in spherical-altitude
94
coordinates simplies to
F
c
M
a
= 2v i
2Ov sin +j
2Ousin (4.34)
Dening the Coriolis parameter as
f = 2Osin (4.35)
gives another form of the Coriolis term as
F
c
M
a
i
f v +j
f u = f
k
r
0 0 1
u v 0
= f k
r
v
h
(4.36)
Equation (4.36) can be approximated in Cartesian coordinates by substituting i, j,
and k for i
, j
, and k
r
. The magnitude of (4.36) is |F
c
|M
a
= f |v
h
| = f
u
2
+v
2
,
where |v
h
| is the horizontal wind speed (m s
1
).
Example 4.4
A mean wind speed of |v
h
| = 10 m s
1
at the North Pole results in a Coriolis
acceleration magnitude of about |F
c
|M
a
= 0.001454 m s
2
.
4.2.3 Gravitational force
Gravity is a real force that acts on a parcel of air. The gravity that we experience is
really a combination of true gravitational force and the Earths apparent centrifu-
gal force. True gravitational force acts toward the center of the Earth. The Earths
apparent centrifugal force, which acts away from the axis of rotation of the Earth,
slightly displaces the direction and magnitude of the true gravitational force. The
sum of the true gravitational and apparent centrifugal force vectors gives an effec-
tive gravitational force vector, which acts normal to the surface of the Earth but
not toward its center.
The Earths apparent centrifugal force (or centripetal acceleration) arises because
the Earth rotates. To an observer xed in space, objects moving with the surface of
a rotating Earth exhibit an inward centripetal acceleration. The object, itself, feels
as if it is being pushed outward, during rotation, by an apparent centrifugal force.
The force is the greatest at the Equator, where the component of the Earths angular
velocity normal to the Earths axis of rotation is the greatest, and zero at the poles,
where the normal component of the Earths angular velocity is zero. Over time,
the apparent centrifugal force has caused the Earth to bulge at the Equator and
compress at the poles. The equatorial radius of Earth is now about 21 km longer
than the polar radius, making the Earth an oblate spheroid.
95
j
R
e
R
e
R
e
cos j
*
F
g
F
r
k
r
k
r
*
F
g
Figure 4.6 Gravitational force
components for the Earth. True
gravitational force acts toward
the center of the Earth, and
apparent centrifugal force acts
away from its axis of rota-
tion. The effective gravitational
force, which is the sum of the
true gravitational and appar-
ent centrifugal forces, acts nor-
mal to the true surface but
not toward the center of the
Earth. The apparent centrifu-
gal force (negative centripetal
acceleration) has caused the
Earth to bulge at the Equa-
tor, as shown in the diagram,
making the Earth an oblate
spheroid. Vectors correspond-
ing to a true sphere are marked
with asterisks to distinguish
them from those corresponding
to the oblate spheroid.
The true gravitational force vector per unit mass, which acts toward the center
of Earth, is
F
g
M
a
= k
r
g
(4.37)
where g
is the true gravitational acceleration. The vectors i
, j
, and k
r
are unit
vectors on a true sphere. The vectors i
, j
, and k
r
are unit vectors on the Earth,
which is an oblate spheroid. Figure 4.6 shows the orientation of true gravitational
force and vertical unit vectors for Earth and for a perfect sphere.
True gravitational acceleration is derived from Newtons law of universal grav-
itation. This law gives the gravitational force vector between two bodies as
F
12.g
= r
21
GM
1
M
2
r
3
21
(4.38)
96
where G is the universal gravitational constant (6.6720 10
11
m
3
kg
1
s
2
),
M
1
and M
2
are the masses of the two bodies, respectively, F
12.g
is the vector force
exerted on M
2
by M
1
, r
21
is the distance vector pointing from body 2 to body 1,
r
21
is the distance between the centers of the two bodies, and the negative sign
indicates that the force acts in a direction opposite to that of the vector r
21
. The
magnitude of the gravitational force vector is F
g
= GM
1
M
2
r
2
21
.
In the case of the Earth, F
12.g
= F
g
, r
21
= R
e
= k
r
R
e
, M
1
= M
a
, M
2
= M
e
, and
r
21
= R
e
, where M
a
is the mass of a parcel of air, and M
e
is the mass of the Earth.
Substituting these values into (4.38) gives the true gravitational force vector per
unit mass as
F
g
M
a
= k
r
GM
e
R
2
e
(4.39)
Equating (4.39) with (4.37) and taking the magnitude give
|F
g
|
M
a
= g
=
GM
e
R
2
e
(4.40)
Example 4.5
The mass of the Earth is about M
e
= 5.98 10
24
kg, and the mean radius is
R
e
= 6.37 10
6
m. Thus, from (4.40), the true gravitational acceleration has a
magnitude of about g
= 9.833 m s
2
.
For a true sphere, the apparent centrifugal force per unit mass expands to
F
r
M
a
= a
r
= (R
e
) = O
r
0 cos sin
R
e
cos 0 0
= j
R
e
O
2
cos sin +k
r
R
e
O
2
cos
2
(4.41)
where
= j
Ocos +k
r
Osin R
e
= k
r
R
e
(4.42)
are the angular velocity vector and radius vector, respectively, of the Earth as if
it were a true sphere, R
e
cos is the perpendicular distance between the axis of
rotation of Earth and the surface of Earth at latitude , as shown in Fig. 4.6,
and
R
e
= O
0 cos sin
0 0 R
e
= i
R
e
Ocos (4.43)
97
The magnitude of (4.41) is |F
r
|M
a
= R
e
O
2
cos . Adding (4.41) to (4.37) gives the
effective gravitational force vector per unit mass on the Earth as
F
g
M
a
=
F
g
M
a
+
F
r
M
a
= j
R
e
O
2
cos sin +k
r
(R
e
O
2
cos
2
g
) = k
r
g (4.44)
where k
r
is the unit vector normal to the oblate spheroid surface of the Earth, and
g = [(R
e
O
2
cos sin)
2
+(g
R
e
O
2
cos
2
)
2
]
12
(4.45)
is the magnitude of the gravitational force per unit mass, or effective gravitational
acceleration (effective gravity). The effective gravity at sea level varies from g =
9.799 m s
2
at the Equator to g = 9.833 m s
2
at the poles. These values are
much larger than accelerations due to the Coriolis effect. The effective gravity
at the poles equals the true gravitational acceleration there, since the apparent
centrifugal acceleration does not act at the poles. Centripetal acceleration affects
true gravitational acceleration by about 0.34 percent at the Equator. The difference
between the equatorial and polar radii of Earth is about 21 km, or 0.33 percent of
an average Earths radius. Thus, apparent centrifugal force appears to account for
the bulging of the Earth at its Equator.
The globally averaged effective gravity at the Earths topographical surface,
which averages 231.4 mabove sea level, is approximately g
0
=9.8060 ms
2
. Since
the variation of the effective gravity g with latitude is small, g is often approximated
with g
0
in models of the Earths lower atmosphere.
Figure 2.1(c) and Appendix Table B.1 give the globally averaged effective gravity
versus altitude. The gure and data were derived by replacing R
e
with R
e
+ z in
(4.40), combining the result with (4.45), and averaging the result globally.
Example 4.6
Both g
and g vary with altitude in the Earths atmosphere. Equation (4.40)

predicts that, 100 km above the equator, g
9.531 m s
2
, or 3.1 percent lower
than its surface value. Equation (4.45) predicts that, at 100 km, g 9.497 ms
2
,
also 3.1 percent lower than its surface value. Thus, the variation of gravity with
altitude is more signicant than is the variation of gravity due to centripetal
acceleration (0.34 percent).
Effective gravity is used to calculate geopotential, the work done against gravity
to raise a unit mass of air from sea level to a given altitude. Geopotential is a scalar
that is a measure of the gravitational potential energy of air per unit mass. The
magnitude of geopotential (m
2
s
2
) is
+(z) =
z
0
g (z) dz (4.46)
98
F
p,r
F
p,l
x
y
z
Figure 4.7 Example of pressure-
gradient forces acting on both sides
of a parcel of air. F
p,r
is the force act-
ing on the right side, and F
p,l
is the
force acting on the left side.
where z = 0 corresponds to sea level and g(z) is gravitational acceleration as a
function of height above the Earths surface (see Appendix Table B.1).
Geopotential height is dened as
Z=
+(z)
g
0
(4.47)
Near the Earths surface, geopotential height approximately equals altitude (Z z)
since g(z) g
0
. Geopotential height differs from actual altitude by about 1.55
percent at 100 km. At 25 km, the difference is 0.39 percent. The assumptions
Z z and g(z) g
0
= g are often made in models of the lowest 100 km of the
atmosphere. Under these assumptions, the magnitude and gradient of geopotential
height are
+(z) gz +(z) = + = k
r
+(z)
z
k
r
g (4.48)
respectively. Substituting (4.48) into (4.44) gives the effective gravitational force
per unit mass in spherical coordinates as
F
g
M
a
= k
r
g = + (4.49)
This equation can be written in Cartesian-altitude coordinates by substituting k
for k
r
.
4.2.4 Pressure-gradient force
The pressure-gradient force is a real force that causes air to move from regions of
high pressure to regions of lowpressure. The force results frompressure differences.
Suppose a cubic parcel of air has volume LxLyLz, as shown in Fig. 4.7. Suppose
99
L
1012 hPa 1008 hPa
100 km
H
Figure 4.8 Example of a pressure
gradient. The difference in pressure
over a 100-kmdistance is 4 hPa. The
letters H and L indicate high and
low pressure, respectively. The thick
arrow indicates the direction of the
pressure-gradient force.
also that air pressures on the right and left sides of the parcel impart the forces
F
p.r
=
p
c
+
p
x
Lx
2
LyLz F
p.l
=
p
c
p
x
Lx
2
LyLz (4.50)
respectively, where p
c
is the pressure at the center of the parcel. Dividing the sum
of these forces by the mass of the parcel, M
a
=
a
LxLyLz, and allowing Lx, Ly,
and Lz to approach zero give the pressure-gradient force per unit mass in the
x-direction as
F
p.x
M
a
=
1
a
p
a
x
(4.51)
Example 4.7
Figure 4.8 shows two isobars, or lines of constant pressure, 100 km apart. The
pressure difference between the isobars is 4 hPa. Assuming
a
= 1.2 kg m
3
,
the magnitude of the horizontal pressure-gradient force per unit mass is
approximately
1
a
p
a
x

1
1.2kg m
3
1012 1008 hPa

10
5
m
100 kg m
1
s
2
hPa
= 0.0033 ms
2
which is much smaller than the force per unit mass due to gravity, but on the
same scale as the Coriolis force per unit mass.
The pressure-gradient force per unit mass can be generalized for three directions
in Cartesian-altitude coordinates with
F
p
M
a
=
1
a
p
a
=
1
i
p
a
x
+j
p
a
y
+k
p
a
z
(4.52)
100
and in spherical-altitude coordinates with
F
p
M
a
=
1
a
p
a
=
1
1
R
e
cos
p
a
e
+j
1
R
e
p
a
+k
r
p
a
z
(4.53)
Example 4.8
In the vertical, the pressure-gradient force per unit mass is much larger than is
that in the horizontal. Pressures at sea level and 100-m altitude are p
a
1013
and 1000 hPa, respectively. The average air density at 50 m is
a
1.2 kg m
3
.
In this case, the pressure-gradient force per unit mass in the vertical is
1
a
p
a
z

1
1.2kg m
3
1013 1000 hPa

100 m
100 kg m
1
s
2
hPa
= 10.8ms
2
which is over 3000 times greater than the horizontal pressure-gradient force
per unit mass calculated in Example 4.7.
4.2.5 Viscous force
Molecular viscosity is a property of a uid that increases its resistance to motion.
This resistance arises for a different reason in liquids than in gases. In liquids,
viscosity is an internal friction that arises when molecules collide with each other
and briey bond, for example by hydrogen bonding, van der Waals forces, or
attractive electric charge. Kinetic energy of the molecules is converted to energy
required to break the bonds, slowing the ow of the liquid. At high temperatures,
bonds between liquid molecules are easier to break, thus viscosity is lower and
liquids move more freely than at low temperatures.
In gases, viscosity is the transfer of momentum between colliding molecules. As
gas molecules collide, they generally do not bond, so there is little net loss of energy.
When a fast molecule collides with a slow molecule, the faster molecule is slowed
down, the slower molecule increases in speed, and both molecules are redirected.
As a result, viscosity reduces the spread of a plume of gas. As temperatures increase,
the viscosity of a gas increases because high temperatures increase the kinetic energy
of each gas molecule, thereby increasing the probability that a molecule will collide
with and exchange momentum with another gas molecule. Due to the increase in
resistance to ow due to enhanced collision and redirection experienced by a gas
at high temperature, a ame burning in air, for example, will not rise so high at a
high temperature as it will at a low temperature. In sum, the effect of heating on
viscosity is opposite for gases than for liquids.
When gas molecules collide with a stationary surface, they impart momentum
to the surface, causing a net loss of energy among gas molecules. The dissipation
of kinetic energy contained in gas molecules at the ground due to viscosity causes
the wind speed at the ground to be zero and to increase logarithmically above the
101
ground. In the absence of molecular viscosity, the wind speed immediately above
the ground would be large.
The molecular viscosity of air can be quantied with
a
=
5
16Ad
2
a
m
a
R
1.8325 10
5
416.16
T +120
T
296.16
1.5
(4.54)
(kg m
1
s
1
), called the dynamic viscosity of air. The rst expression is based on
gas kinetic theory and can be extended to any gas, and the second expression is
based on experiment and referred to as Sutherlands equation (List 1984). In the
equations, m
a
is the molecular weight of air (28.966 g mol
1
), R
is the univer-
sal gas constant (8314.51 g m
2
s
2
mol
1
K
1
), T is absolute temperature (K), A
is Avogadros number (molec. mol
1
), and d
a
is the average diameter of an air
molecule (m). Equating the two expressions at 296.16 K gives d
a
3.673
10
10
m. A related parameter is the kinematic viscosity of air
a
=

a
a
(4.55)
(m
2
s
1
), which is a molecular diffusion coefcient for air, analogous to the molec-
ular diffusion coefcient for a trace gas.
Viscous interactions among air molecules sliding over each other give rise to a
viscous force, which is an internal force caused by molecular interactions within a
parcel. The change of wind speed with height (i.e., uz) is wind shear. As layers
of air slide over one another at different speeds due to wind shear, each layer exerts
a viscous stress (shearing stress), or force per unit area, on the other. The stress
acts parallel to the direction of motion and over a plane normal to the direction of
shear. If wind shear in the z-direction exerts a force in the x-direction per unit area
of the xy plane, the resulting shearing stress is
zx
=
a
u
z
(4.56)
(N m
2
or kg m
1
s
2
), where
a
is the dynamic viscosity of air from (4.54) and is
also the ratio of shearing stress to shear. Shearing stress results when momentum
is transported down a gradient of velocity, just as gas molecules are transported
by molecular diffusion down a gradient of gas concentration. A cubic parcel of air
experiences a shearing stress on its top and bottom, as shown in Fig. 4.9.
The net viscous force on a parcel of air equals the shearing stress on the top
minus the shearing stress on the bottom, multiplied by the area over which the
stress acts. The force acts parallel to the direction of motion. If
zx
is the shearing
stress in the middle of the parcel, and if
zx
z is the vertical gradient of shearing
stress, the shearing stresses at the top and bottom, respectively, of the parcel are
zx.top
=
zx
+

zx
z
Lz
2

zx.bot
=
zx

zx
z
Lz
2
(4.57)
102
x
y
z
zx, top
zx, bot
zx, mid
Figure 4.9 Example of shearing
stress in the x-direction on a volume
of air.
u
z
Net viscous force
zx, mid
u
z
zx, bot
zx, top
No net viscous force
x x
zx, mid
zx, bot
zx, top
Figure 4.10 A linear vertical wind shear results in a constant shearing
stress at all heights and no net viscous force. A nonlinear wind shear
results in a change of shearing stress with height and a net viscous force.
Subtracting the shearing stress at the bottom from that at the top of the parcel,
multiplying by area, and dividing by air parcel mass, M
a
=
a
LxLyLz, give the
net viscous force per unit mass as
F
v.zx
M
a
= (
zx.top
zx.bot
)
LxLy
a
LxLyLz
=
1
zx
z
(4.58)
Substituting shearing stress from (4.57) into (4.58), and assuming
a
is invariant
with altitude, give
F
v.zx
M
a
=
1
a
u
z
2
u
z
2
(4.59)
This equation suggests that, if wind speed does not change with height or changes
linearly with height and
a
is constant, the viscous force per unit mass in the x-
direction due to shear in the z-direction is zero. In such cases, shearing stress is
constant with height, and no net viscous force occurs. When wind speed changes
nonlinearly with height, the shearing stress at the top of a parcel differs from that
at the bottom, and the net viscous force is nonzero. Figure 4.10 illustrates these
two cases.
Expanding (4.59) gives the viscous-force vector per unit mass as
F
v
M
a
=

a
2
v =
a
2
v (4.60)
103
where
a
was given in (4.55). The gradient squared term expands in Cartesian-
altitude coordinates to
2
v = (
) v = i
2
u
x
2
+

2
u
y
2
+

2
u
z
2
+ j
2
v
x
2
+

2
v
y
2
+

2
v
z
2
+k
2
w
x
2
+

2
w
y
2
+

2
w
z
2
(4.61)
Substituting (4.1) and spherical coordinate unit vectors converts (4.61) to spherical-
altitude coordinates.
Viscous forces in the atmosphere are small, except adjacent to a surface, where
collision of molecules with the surface results in a loss of kinetic energy to the
surface. The loss of momentumat the surface causes winds to be zero at the surface
but to increase logarithmically immediately above the surface. The relatively large,
nonlinear wind-speed variation over a short distance increases the magnitude of
the viscous force.
Example 4.9
Viscous forces away from the ground are small. Suppose u
1
= 10 m s
1
at
altitude z
1
= 1000 m, u
2
= 14 m s
1
at z
2
= 1250 m, and u
3
= 20 m s
1
at
z
3
=1500 m. If the average temperature and air density at z
2
are T =280 K and
a
= 1.085 kg m
3
, respectively, the dynamic viscosity of air is
a
= 1.753
10
5
kg m
1
s
1
. The resulting viscous force per unit mass in the x-direction
due to wind shear in the z-direction is approximately
F
v.zx
M
a
a
1
(z
3
z
1
)2
u
3
u
2
z
3
z
2
u
2
u
1
z
2
z
1
= 5.17 10
10
ms
2
which is much smaller than the pressure-gradient force per unit mass from
Example 4.7.
Example 4.10
Viscous forces near the ground are often signicant. Suppose u
1
= 0 m s
1
at
altitude z
1
= 0 m, u
2
= 0.4 m s
1
at z
2
= 0.05 m, and u
3
= 1 m s
1
at z
3
=
0.1 m. If the average temperature and air density at z
2
are T = 288 K and
a
=
1.225 kg m
3
, the dynamic viscosity of air is
a
= 1.792 10
5
kg m
1
s
1
.
The resulting viscous force per unit mass in the x-direction due to wind shear
in the z-direction is
F
v.zx
M
a
a
1
(z
3
z
1
)2
u
3
u
2
z
3
z
2
u
2
u
1
z
2
z
1
= 1.17 10
3
ms
2
which is comparable in magnitude with the horizontal pressure-gradient force
per unit mass calculated in Example 4.7.
104
4.2.6 Turbulent-ux divergence
At the ground, wind speeds are zero. In the surface layer, which is a 50300-
m-thick region of the atmosphere adjacent to the surface, wind speeds increase
logarithmically with increasing height, creating wind shear. Wind shear produces
a shearing stress that enhances collisions on a molecular scale. On larger scales,
wind shear produces rotating air motions, or eddies. Wind shear can arise when,
for example, moving air encounters an obstacle, such as a rock, tree, structure, or
mountain. The obstacle slows down the wind beyond it, but the wind above the
obstacle is still fast, giving rise to a vertical gradient in wind speed that produces
a rotating eddy. Eddies created downwind of obstacles are called turbulent wakes.
Eddies can range in size from a few millimeters in diameter to the size of the
boundary layer (several hundred meters in diameter).
Eddies are created not only by wind shear but also by buoyancy. Heating of sur-
face air causes the air to rise in a thermal, forcing air aloft to sink to replace the rising
air, creating a circulation system. Turbulence consists of many eddies of different
size acting together. Wind shear arising from obstacles is called mechanical shear,
and the resulting turbulence is called mechanical turbulence. Thermal turbulence
is turbulence due to eddies of different size arising from buoyancy. Mechanical
turbulence is most important in the surface layer, and thermal turbulence is most
important in the mixed layer. Thermal turbulence magnies the effect of mechan-
ical turbulence by enabling eddies to extend to greater heights, increasing their
ability to exchange air between the surface and the mixed layer (Stull 1988).
Because eddies are circulations of air, they transfer momentum, energy, gases,
and particles vertically and horizontally. For example, eddies transfer fast winds
aloft toward the surface, creating gusts near the surface, and slow winds near the
surface, aloft. Both effects reduce wind shear. The transfer of momentum by eddies
is analogous to the smaller-scale transfer of momentum by molecular viscosity. As
such, turbulence due to eddies is often referred to as eddy viscosity.
If a numerical model has grid-cell resolution on the order of a few millimeters,
it resolves the ow of eddies of all sizes without the need to treat such eddies as
subgrid phenomena. Such models, called direct numerical simulation (DNS) models
(Section 8.4) cannot ignore the viscous force because, at that resolution, the viscous
force is important. All other models must parameterize the subgrid eddies to some
degree (Section 8.4).
In models that parameterize subgrid-scale eddies, the parameterization can be
obtained as follows. Multiplying the precise acceleration of air from (4.22) by
a
,
multiplying the continuity equation for air from (3.20) by v, and adding the results
yield
a
a
l
=
a
v
t
+(v
) v
+v
a
t
+
(v
a
)
(4.62)
The variables in this equation are decomposed as v = v +v
and
a
=
a
+
a
.
Because density perturbations are generally small (
a

a
), the density simplies
105
to
a

a
. Substituting the decomposed values into (4.62) gives

a
a
l
=
a
( v +v
)
t
+[( v +v
]( v +v
+( v +v

a
t
+
[( v +v
)
a
]
(4.63)
Taking the time and grid-volume average of this equation, eliminating zero-value
terms, and removing unnecessary overbars give

a
a
l
=
a
v
t
+( v
) v
+ v

a
t
+
(v
a
)
+
a
(v
) v
+v
(v

a
)
(4.64)
Substituting the time- and grid-volume-averaged continuity equation for air from
(3.48) into (4.64) and dividing through by
a
yield a
l
= a
l
+a
l
,where
a
l
=
v
t
+( v
) v a
l
=
F
t
M
a
=
1

a

a
(v
) v
+v
(v

a
)
(4.65)
The second term is treated as a force per unit mass in the momentum equation,
where F
t
is a turbulent-ux divergence vector multiplied by air mass. Since F
t
originates on the left side of (4.12), it must be subtracted from the right side when
treated as a force. F
t
accounts for mechanical shear, buoyancy, and other eddy
effects, such as waves. Expanding the turbulent-ux divergence term in (4.65) in
Cartesian-altitude coordinates gives
F
t
M
a
= i
1
(
a
u
)
x
+
(
a
v
)
y
+
(
a
w
)
z
+j
1
(
a
u
)
x
+
(
a
v
)
y
+
(
a
w
)
z
+k
1
(
a
u
)
x
+
(
a
v
)
y
+
(
a
w
)
z
(4.66)
Averages, such as u
and w
(m m s
1
s
1
), are kinematic turbulent uxes of
momentum, since they have units of momentumux (kg ms
1
m
2
s
1
) divided by
air density (kg m
3
). Each such average must be parameterized. A parameteriza-
tion introduced previously and discussed in Chapter 8 is K-theory. With K-theory,
vertical kinematic turbulent uxes of westeast and southnorth momentum are
approximated with
w
= K
m.zx
u
z
w
= K
m.zy
v
z
(4.67)
respectively, where the K
m
s are eddy diffusion coefcients for momentum (m
2
s
1
or cm
2
s
1
). In all, nine eddy diffusion coefcients for momentumare needed. Only
three were required for energy. Since u
= v
, u
= w
, and v
= w
, the
106
number of coefcients for momentum can be reduced to six. Substituting (4.67)
and other like terms into (4.66) and dropping overbars for simplicity give the
turbulent-ux divergence as
F
t
M
a
= i
1
a
K
m.xx
u
x
+

y
a
K
m.yx
u
y
+

z
a
K
m.zx
u
z
j
1
a
K
m.xy
v
x
+

y
a
K
m.yy
v
y
+

z
a
K
m.zy
v
z
k
1
a
K
m.xz
w
x
+

y
a
K
m.yz
w
y
+

z
a
K
m.zz
w
z
(4.68)
This equation can be transformed to spherical-altitude coordinates by substituting
(4.1) and spherical-coordinate unit vectors into it. Each termin (4.68) represents an
acceleration in one direction due to transport of momentum normal to that direc-
tion. For example, the zx termis an acceleration in the x-direction due to a gradient
in wind shear and transport of momentum in the z-direction. Kinematic turbulent
uxes are analogous to shearing stresses in that both result when momentum is
transported down a gradient of velocity. In the case of viscosity, momentum is
transported by molecular diffusion. In the case of turbulence, momentum is trans-
ported by eddy diffusion. As with the viscous forces per unit mass in (4.60), the
turbulent-ux divergence terms in (4.68) equal zero when the
a
K
m
terms are con-
stant in space and the wind shear is either zero or changes linearly with distance.
Usually, the
a
K
m
terms vary in space.
The eddy diffusion coefcients in (4.68) can be written in tensor form as
K
m
=
K
m.xx
0 0
0 K
m.yx
0
0 0 K
m.zx
for u,
K
m.xy
0 0
0 K
m.yy
0
0 0 K
m.zy
for v,
K
m.xz
0 0
0 K
m.yz
0
0 0 K
m.zz
for w (4.69)
where a different tensor is used depending on whether the u, v, or w momentum
equation is being solved. In vector and tensor notation, (4.68) simplies to
F
t
M
a
=
1
a
(
a
K
m
) v (4.70)
where the choice of tensor K
m
depends on whether the scalar in vector v is u, v,
or w.
107
Example 4.11
What is the westeast acceleration of wind at height 300 m due to the
downward transfer of westerly momentum from 400 m under the following
conditions: u
1
= 10 m s
1
at altitude z
1
= 300 m, u
2
= 12 m s
1
at z
2
= 350
m, and u
3
= 15 m s
1
at z
3
= 400 m? Assume also that a typical value of K
m
in the vertical direction in the middle of the boundary layer is 50 m
2
s
1
and
that the diffusion coefcient and air density remain constant between 300 and
400 m.
SOLUTION
F
t.zx
M
a
=
1
a
K
m.zx
u
z
K
m.zx
(z
3
z
1
)2
u
3
u
2
z
3
z
2
u
2
u
1
z
2
z
1
= 0.02 ms
2
Example 4.12
What is the deceleration of westerly wind in the south due to transfer of
momentum to the north under the following conditions: u
1
= 10 m s
1
at
location y
1
=0 m, u
2
=9 m s
1
at y
2
=500 m, and u
3
=7 m s
1
at y
3
=1000 m?
Assume also that a typical value of K
m
in the horizontal is about K
m
= 100 m
2
s
1
and that the diffusion coefcient and air density remain constant between
y
1
and y
3
.
SOLUTION
F
t.yx
M
a
=
1
a
K
m.yx
u
y
K
m.yx
(y
3
y
1
)2
u
3
u
2
y
3
y
2
u
2
u
1
y
2
y
1
= 0.0004 ms
2
4.2.7 Complete momentum equation
Table 4.1 summarizes the accelerations and forces per unit mass derived above and
gives approximate magnitudes of the terms. Noting that a
l
= a
l
+F
t
M
a
, removing
overbars for simplicity, and substituting terms from Table 4.1 into (4.21) give the
vector form of the momentum equation as
dv
dt
= f k v +
1
a
p
a
+

a
2
v +
1
a
(
a
K
m
) v (4.71)
108
Table 4.1 Terms in the momentum equation and their horizontal
and vertical magnitudes
Acceleration or Horizontal Vertical
Term force/mass expression acceleration (m s
2
) acceleration (m s
2
)
Local acceleration a
l
=
d v
dt
=
v
t
+( v
) v 10
4 a
10
7
1
Coriolis force per
unit mass
F
c
M
a
= f k v 10
3
0
Effective
gravitational force
per unit mass
F
g
M
a
=
F
g
M
a
+
F
r
M
a
= + 0 10
Pressure-gradient
force per unit mass
F
p
M
a
=
1
a
p
a
10
3
10
Viscous force per unit
mass
F
v
M
a
=

a
2
v
b
10
12
10
3 b
10
15
10
5
Turbulent-ux
divergence of
momentum
F
t
M
a
=
1
a
(
a
K
m
) v
c
00.005
c
01
a
Low value for large-scale motions, high value for small-scale motions (-3 km).
b
Low value for free atmosphere, high value for air adjacent to the surface.
c
Low value for no wind shear, high value for large wind shear.
Table 4.1 shows that some terms in these equations are unimportant, depending
on the scale of motion. Three dimensionless parameters the Ekman number,
Rossby number, and Froude number are used for scale analysis to estimate the
importance of different processes. These parameters are
Ek =

a
ux
2
uf
Ro =
u
2
x
uf
Fr
2
=
w
2
z
g
(4.72)
respectively. The Ekman number gives the ratio of the viscous force to the Coriolis
force. Above the ground, viscous terms are unimportant relative to Coriolis terms.
Thus, the Ekman number is small, allowing the viscous termto be removed fromthe
momentum equation without much loss in accuracy. The Rossby number gives the
ratio of the local acceleration to the Coriolis force per unit mass. Local accelerations
are more important than viscous accelerations, but less important than Coriolis
accelerations. Thus, the Rossby number is much larger than the Ekman number,
but usually less than unity. The Froude number gives the ratio of local acceleration
to gravitational acceleration in the vertical. Over large horizontal scales (>3 km),
vertical accelerations are small in comparison with gravitational accelerations. In
such cases, the vertical acceleration term is often removed from the momentum
equation, resulting in the hydrostatic assumption.
109
Example 4.13
Over large horizontal scales
a
10
6
m
2
s
1
, u 10 m s
1
, x 10
6
m,
f 10
4
s
1
, n 0.01 m s
1
, and z 10
4
m. The Ekman number under these
conditions is Ek = 10
14
, indicating that viscous forces are small. The Rossby
number has a value of about Ro =0.1, indicating that local accelerations are an
order of magnitude smaller than Coriolis accelerations. The Froude number is
Fr = 3 10
5
. Thus, local accelerations in the vertical are unimportant over
large horizontal scales.
Since air viscosity is negligible for most atmospheric scales, it can be ignored
in the momentum equation. Removing viscosity from (4.71) and expanding the
equation in Cartesian-altitude coordinates give
du
dt
=
u
t
+u
u
x
+v
u
y
+w
u
z
= f v
1
a
p
a
x
+
1
a
K
m.xx
u
x
+

y
a
K
m.yx
u
y
+

z
a
K
m.zx
u
z
(4.73)
dv
dt
=
v
t
+u
v
x
+v
v
y
+w
v
z
= f u
1
a
p
a
y
+
1
a
K
m.xy
v
x
+

y
a
K
m.yy
v
y
+

z
a
K
m.zy
v
z
(4.74)
dw
dt
=
w
t
+u
w
x
+v
w
y
+w
w
z
= g
1
a
p
a
z
+
1
a
K
m.xz
w
x
+

y
a
K
m.yz
w
y
+

z
a
K
m.zz
w
z
(4.75)
The momentum equation in spherical-altitude coordinates requires additional
terms. When a reference frame xed on the surface of the Earth (noninertial ref-
erence frame) is used, the spherical-altitude coordinate conversion terms from
(4.30), uv tanR
e
and u
2
tanR
e
, should be treated as apparent forces and are
110
4.3 Applications of the momentum equation
moved to the right side of the momentum equation. Implementing this change and
substituting (4.1) into (4.73)(4.75) give approximate forms of the directional
momentum equations in spherical-altitude coordinates as
u
t
+
u
R
e
cos
u
e
+
v
R
e
u
+w
u
z
=
uv tan
R
e
+ f v
1
a
R
e
cos
p
a
e
+
1
1
R
2
e
cos
a
K
m.xx
cos
u
+
1
R
2
e
a
K
m.yx
u
+

z
a
K
m.zx
u
z
(4.76)
v
t
+
u
R
e
cos
v
e
+
v
R
e
v
+w
v
z
=
u
2
tan
R
e
f u
1
a
R
e
p
a
+
1
1
R
2
e
cos
a
K
m.xy
cos
v
+
1
R
2
e
a
K
m.yy
v
+

z
a
K
m.zy
v
z
(4.77)
w
t
+
u
R
e
cos
w
e
+
v
R
e
w
+w
w
z
= g
1
a
p
a
z
+
1
1
R
2
e
cos
a
K
m.xz
cos
w
+
1
R
2
e
a
K
m.yz
w
+

z
a
K
m.zz
w
z
(4.78)
4.3 APPLICATIONS OF THE MOMENTUM EQUATION
The continuity equation for air, the species continuity equation, the thermodynamic
energy equation, the three momentum equations (one for each direction), and the
equation of state are referred to here as the equations of atmospheric dynamics.
Removing the species continuity equation from this list and substituting the hydro-
static equation in place of the full vertical momentum equation give the primitive
equations, which are a basic form of the Eulerian equations of uid motion. Many
atmospheric motions, including the geostrophic wind, surface winds, the gradi-
ent wind, surface winds around high- and low- pressure centers, and atmospheric
111
H
L
Surface
H
L
F
p
Aloft
F
p
F
c
F
c
F
t
F
t
+F
c
v
v
Figure 4.11 Force and wind vectors aloft and at the
surface in the Northern Hemisphere. The parallel
lines are isobars.
waves, can be understood by looking at simplied forms of the equations of atmo-
spheric dynamics.
4.3.1 Geostrophic wind
The momentum equation can be simplied to isolate air motions, such as the
geostrophic wind. The geostrophic (Earth-turning) wind arises when the Coriolis
force exactly balances the pressure-gradient force. Such a balance occurs following
a geostrophic adjustment process, described in Fig. 4.19 (Section 4.3.5.3). Many
motions in the free troposphere are close to being geostrophic because, in the free
troposphere, horizontal local accelerations and turbulent accelerations are usually
much smaller than Coriolis and pressure-gradient accelerations. Removing local
acceleration and turbulence terms from (4.73) and (4.74), and solving give the
geostrophic scalar velocities in Cartesian-altitude coordinates as
v
g
=
1
f
a
p
a
x
u
g
=
1
f
a
p
a
y
(4.79)
In geostrophic equilibrium, the pressure-gradient force is equal in magnitude to and
opposite in direction to the Coriolis force. The resulting geostrophic wind ows
90
to the left of the Coriolis force in the Northern Hemisphere, as shown in the
top portion of Fig. 4.11. In the Southern Hemisphere, the geostrophic wind ows
90
to the right of the Coriolis force.

112
Example 4.14
If =30
N,
a
=0.76 kg m
3
, and the pressure gradient is 4 hPa per 150 kmin
the southnorth direction, estimate the westeast geostrophic wind speed.
SOLUTION
From (4.35), f = 7.292 10
5
rads
1
. From (4.79), u
g
= 48.1 m s
1
.
In vector form, the geostrophic velocity in Cartesian-altitude coordinates is
v
g
= iu
g
+jv
g
=
1
f
a
i
p
a
y
+j
p
a
x
=
1
f
a
i j k
0 0 1
p
a
x
p
a
y
0
=
1
f
a
k
z
p
a
(4.80)
where
z
=
i

x
z
+
j

y
z
= i

x
+j

y
(4.81)
is the horizontal gradient operator in Cartesian-altitude coordinates. The subscript
z indicates that the partial derivative is taken along a surface of constant altitude;
thus, kz = 0. Equation (4.80) indicates that the geostrophic wind ows parallel
to lines of constant pressure (isobars).
4.3.2 Surface-layer winds
In steady state, winds in the surface layer are affected primarily by the pressure-
gradient force, Coriolis force, mechanical turbulence, and thermal turbulence.
Objects protruding from the surface, such as blades of grass, bushes, trees, struc-
tures, hills, and mountains, slowwinds near the surface. Aloft, fewer obstacles exist,
and wind speeds are generally higher than at the surface. In terms of the momen-
tum equation, forces that slow winds near the surface appear in the turbulent-ux
divergence vector F
t
.
Near the surface and in steady state, the sum of the Coriolis force and turbu-
lent ux vectors balance the pressure-gradient force vector. Aloft, the Coriolis and
pressure-gradient force vectors balance each other. Figure 4.11 shows forces and
resulting winds in the Northern Hemisphere in both cases. On average, friction near
the surface shifts winds about 30
counterclockwise in comparison with winds aloft

in the Northern Hemisphere, and 30
clockwise in comparison with winds aloft

in the Southern Hemisphere. The variation in wind direction may be 45
or more,
depending on the roughness of the surface. In both hemispheres, surface winds are
tilted toward low pressure.
113
Cloud layer
Entrainment zone
Inversion layer
Free troposphere
Neutral
convective
mixed layer
Surface layer
Subcloud layer
Daytime mean wind speed
A
l
t
i
t
u
d
e
B
o
u
n
d
a
r
y

l
a
y
e
r
Entrainment zone
Inversion layer
Free troposphere
Surface layer
Nighttime mean wind speed
Residual layer
Stable
boundary
layer
Nocturnal jet
A
l
t
i
t
u
d
e
B
o
u
n
d
a
r
y

l
a
y
e
r
(a) (b)
Figure 4.12 Variation of wind speed with height during the (a) day and
(b) night in the atmospheric boundary layer. Adapted from Stull (1988).
In the presence of near-surface turbulence, the steady-state horizontal momen-
tum equations simplify in Cartesian-altitude coordinates from (4.73)(4.74) to
f v =
1
a
p
a
x
+
1
a
K
m.zx
u
z
f u =
1
a
p
a
y
+
1
a
K
m.zy
v
z
(4.82)
Figure 4.12 shows idealized variations of wind speed with increasing height in
the boundary layer during day and night. During the day, wind speeds increase
logarithmically with height in the surface layer. In the mixed layer, temperature
and wind speeds are relatively uniform with height. Above the entrainment zone,
wind speeds increase to their geostrophic values.
At night, wind speeds near the surface increase logarithmically with increasing
height but are lower than during the day. The reduced mixing in the stable boundary
layer increases wind speed toward the top of the layer, creating a nocturnal or low-
level jet that is faster than the geostrophic wind. In the residual layer, wind speeds
decrease with increasing height and approach the geostrophic wind speed.
Figures 4.13(a) and (b) show measurements of u- and v-scalar velocities at
Riverside, California, during a summer night and day, respectively. Riverside lies
between the coast and a mountain range. At night, the u- and v-winds were small.
In the lower boundary layer, during the day, the u-winds were strong due to a sea
breeze. Aloft, owtoward the ocean was strong due to a thermal lowinland, which
increased pressure and easterly winds aloft inland. The v-component of wind was
relatively small.
4.3.3 The gradient wind
When air rotates around a center of low pressure, as in a midlatitude cyclone,
tornado, or hurricane, or when it rotates around a center of high pressure, as in an
114
15 10 5 0 5 10 15
700
750
800
850
900
950
1000
Wind speed (m s
1
)
P
r
e
s
s
u
r
e

(
h
P
a
)
3:30 a.m.
u v
15 10 5 0 5 10 15
700
750
800
850
900
950
1000
Wind speed (m s
1
)
P
r
e
s
s
u
r
e

(
h
P
a
)
u
v
3:30 p.m.
(a) (b)
Figure 4.13 Measured variation of u and v scalar velocities with height at
(a) 3:30 a.m. and (b) 3:30 p.m. on August 27, 1987, at Riverside, California.
i
j
R
c
jq i
R
O
R
c
q
P
v
q
u
R
a
R
Figure 4.14 Cylindrical coor-
dinate components (i and j are
in Cartesian coordinates).
anticyclone, the horizontal momentum equations are more appropriately written
in cylindrical coordinates.
In Cartesian coordinates, distance variables are x, y, and z, and unit vectors are
i, j, and k. In cylindrical coordinates, distance variables are R
c
, , and z, where R
c
is the radius of curvature, or distance from point O to P in Fig. 4.14, and is the
angle (radians) measured counterclockwise from the positive x-axis. Distances x
and y in Cartesian coordinates are related to R
c
and in cylindrical coordinates
by
x = R
c
cos y = R
c
sin (4.83)
Thus,
R
2
c
= x
2
+ y
2
= tan
1
y
x
(4.84)
115
The unit vectors in cylindrical coordinates are
i
R
= i cos +j sin j
= i sin +j cos k = k (4.85)

which are directed normal and tangential, respectively, to the circle in Fig. 4.14.
The radial vector, directed normal to the circle, is
R
c
= i
R
R
c
(4.86)
The radial velocity, tangential velocity, and angular velocity vectors in cylindrical
coordinates are
u
R
= i
R
u
R
v
= j
w
z
= k
z
(4.87)
respectively, where
u
R
=
dR
c
dt
v
= R
c
d
dt

z
=
d
dt
=
v
R
c
(4.88)
are the radial, tangential, and angular scalar velocities, respectively.
The tangential and angular velocity vectors in cylindrical coordinates are related
to each other by
v
= w
z
R
c
=
i
R
j
k
0 0 v
R
c
R
c
0 0
= j
(4.89)
The local apparent centrifugal force per unit mass, which acts away fromthe center
of curvature, is equal and opposite to the local centripetal acceleration vector,
a
R
=
F
R
M
a
= w
z
v
i
R
j
k
0 0 v
R
c
0 v
= i
R
v
2
R
c
= i
R
a
R
(4.90)
where a
R
= v
2
R
c
is the scalar centripetal acceleration.
In cylindrical coordinates and in the absence of eddy diffusion, the horizontal
momentum equations transform from (4.73) and (4.74) to
du
R
dt
= f v
a
p
a
R
c
+
v
2
R
c
dv
dt
= f u
R
u
R
v
R
c
(4.91)
respectively, where v
2
R
c
and u
R
v
R
c
are centripetal accelerations, treated as
apparent centrifugal forces per unit mass, that arise from the transformation from
Cartesian to cylindrical coordinates. These terms are implicitly included in the
Cartesian-coordinate momentum equations.
When air ows around a center of low or high pressure aloft, as in Fig. 4.15(a)
or (b), respectively, the primary forces acting on the air are the Coriolis, pressure
gradient, and apparent centrifugal forces. If local acceleration is removed from
the rst equation in (4.91), the resulting wind is the gradient wind. Assuming
du
R
dt = 0 in the rst equation in (4.91) and solving for the gradient-wind scalar
116
q
L F
c
F
R
F
p
v
834 hPa
830 hPa
q
H
F
c
F
R
F
p
v
834 hPa
830 hPa
(a) (b)
Figure 4.15 Gradient winds around a center of (a) low and (b) high
pressure in the Northern Hemisphere, and the forces affecting them.
velocity give
v
=
R
c
f
2

R
c
2
f
2
+4
1
R
c
a
p
a
R
c
(4.92)
In this solution, the positive square root is correct and results in v
that is either
positive or negative. The negative root is incorrect and results in an unphysical
solution.
Example 4.15
Suppose the pressure gradient a distance R
c
= 70 km from the center of a
hurricane is p
a
R
c
= 45 hPa per 100 km. Assume = 15
N, p
a
= 850 hPa,
and
a
= 1.06 kg m
3
. These values give :
= 52 m s
1
and :
g
= 1123 ms
1
.
Since :
- :
g
, the apparent centrifugal force slows down the geostrophic
wind.
If the same conditions are used for the high-pressure center case (except
the pressure gradient is reversed), the quadratic has an imaginary root, which
is an unphysical solution. Reducing p
a
R
c
to 0.1 hPa per 100 km in this
example gives :
= 1.7 ms
1
. Thus, the magnitude of pressure gradients and
resulting wind speeds around high-pressure centers are smaller than those
around comparative low-pressure centers.
4.3.4 Surface winds around highs and lows
The deceleration of winds due to drag must be included in the momentumequation
when ow around a low- or high-pressure center near the surface occurs. Figures
4.16(a) and (b) show forces acting on the wind in such cases. In cylindrical coordi-
nates, the horizontal momentum equations describing ow in the presence of the
117
L
F
c
F
R
F
p
v
q
1000 hPa
996 hPa
F
t
1012 hPa
(a) (b)
H
F
c
F
R
F
p
v
q
1016 hPa
F
t
Figure 4.16 Surface winds around centers of (a) low and (b) high pres-
sure in the Northern Hemisphere and the forces affecting them.
Coriolis force, the pressure gradient force, and friction are
du
R
dt
= f v
a
p
a
R
c
+
v
2
R
c
+
1
a
(
a
w
R
)
z
dv
dt
= f u
R
u
R
v
R
c
+
1
a
(
a
w
)
z
(4.93)
where u
R
and v
are the perturbation components of u

R
and v
, respectively.
4.3.5 Atmospheric waves
Atmospheric waves are oscillations in pressure and/or velocity that propagate in
space and time. Because waves perturb air parcels, which contain gas molecules,
aerosol particles, and energy, waves result in spatial and temporal changes in
momentum, concentration, and temperature. Several wave types occur in the atmo-
sphere, including acoustic (sound), gravity (buoyancy), and Rossby (planetary)
waves. Waves are important to understand from a modeling point of view, because
they are often the fastest motion in a model domain and must be ltered from or
treated with special numerical methods in the equations of atmospheric dynamics.
Such ltering is discussed in Section 5.1. In this section, analytical wave equation
solutions are derived from the equations of atmospheric dynamics.
An atmospheric wave consists of a group of individual waves of different charac-
teristics superimposed upon each other. Aset of individual waves can be considered
as a group under the superposition principle, which states that a group of two or
more waves can traverse the same space simultaneously, and the displacement of a
medium due to the group is the sum of the displacements of each individual wave
in the group. As a result of superposition, the shape of the sum of all waves in a
group, called the envelope, differs from the shape of each individual wave in the
group.
Individual waves are transverse, longitudinal, or both. A transverse wave is a
wave that has a sinusoidal wave shape perpendicular to the direction of propagation
118
of the wave. If one end of a spring is xed to a oor and the other end is lifted
vertically and then oscillated horizontally, a sinusoidal transverse wave propagates
down the spring to the oor. Pure gravity waves, light waves, and violin-string
waves are transverse waves. A longitudinal wave is one in which the disturbance
in wave shape ows along (parallel to) the direction of propagation of the wave.
If one end of a spring is xed to the oor and the other end is lifted vertically
and then oscillated vertically, a longitudinal wave propagates to the oor. Pure
acoustic waves are longitudinal. Water waves at the top of the ocean surface are a
combination of transverse and longitudinal waves.
The main characteristics of an individual wave are its wavelength, wavenumber,
frequency of oscillation, phase speed, and amplitude. A wavelength is the distance
(m) between two crests or troughs in a wave. On a Cartesian grid, wavelengths
in the x-, y-, and z-directions are denoted by
.x
,
.y
, and
.z
, respectively. The
wavenumber of a wave is dened here as 2 divided by the wavelength, or the num-
ber of wavelengths in a circle of unit radius. In three dimensions, the wavenumber
vector is
K = i
k +j
l +k m (4.94)
where individual wavenumbers in the x-, y-, and z-directions are related to wave-
length by
k =
2
.x
l =
2
.y
m=
2
.z
(4.95)
(m
1
) respectively. The magnitude of the wavenumber vector is
|
K| =
k
2
+

l
2
+ m
2
(4.96)
The frequency of oscillation (
, s
1
) (the angular frequency in this case) of a wave
is the number of wavelengths that pass through a given point in a circle of unit
radius per unit time. The phase speed (c
, m s
1
) (or wave speed) of a wave is the
speed at which a mathematical surface of constant phase travels. In other words,
it is the speed at which all components of the individual wave travel along the
direction of propagation. The frequency of oscillation is related to the phase speed
and wavenumber by the dispersion relationship,
= c
k
2
+
l
2
+ m
2
= c
K| (4.97)
The amplitude (A
w
, m) of a wave is the magnitude of its maximum displacement.
In the case of transverse waves, the displacement (D, m) is the height of the wave
shape normal to the direction of propagation and varies between A
w
. A common
wave equation for the propagation of a transverse wave is
D(x. t) = A
w
sin
kx
(4.98)
119
0 2 4 6 8 10
2
1
0
1
2
D
i
s
p
l
a
c
e
m
e
n
t

(
m
)
x
Wavelength
(m)
A
m
p
l
i
t
u
d
e
Figure 4.17 Characteristics of a wave
generated from wave equation (4.98)
with A
w
= 1 m, k = 1.745 m
1
, and
.x
= 3.601 m at t = 0.
where the displacement is a function of distance x along the direction of propaga-
tion and time t. The argument in (4.98) is in radians. Figure 4.17 shows character-
istics of a wave generated from (4.98).
When individual waves propagate as a group, the group velocity is the velocity
of the envelope of the group. The three-dimensional group velocity vector is
c
g
= ic
g.x
+jc
g.y
+kc
g.z
(4.99)
where
c
g.x
=

k
c
g.y
=

l
c
g.z
=

m
(4.100)
are group scalar velocities. Substituting (4.97) into (4.100) gives
c
g.x
= c
k
|
K|
+|
K|
c
k
c
g.y
= c
l
|
K|
+|
K|
c
l
c
g.z
= c
m
|
K|
+|
K|
c
m
(4.101)
The magnitude of the group velocity is the group speed,
c
g
c
2
g.x
+c
2
g.y
+c
2
g.z
(4.102)
If waves traveling as a group propagate with a group speed equal to the phase
speed of each wave within the group, the shape of the group does not deform
over time, and the medium in which the group propagates is called a nondispersive
medium. For example, free space is a nondispersive medium for electromagnetic
waves (Chapter 9), and air is a nondispersive medium for sound waves (Section
4.3.5.1). If air were a dispersive mediumfor sound waves, then high-frequency and
low-frequency sounds from a piano would reach a persons ear at different times,
and the resulting sound would not be harmonious.
120
0 0.5 1 1.5 2
D
i
s
p
l
a
c
e
m
e
n
t
Time 1
Time 2
Nondispersive wave
x (m)
0 0.5 1 1.5 2
D
i
s
p
l
a
c
e
m
e
n
t
Time 1
Time 2
Dispersive wave
x (m)
(a) (b)
Figure 4.18 Wave pulses in (a) nondispersive and (b) dispersive media.
In the nondispersive case,
= c
s
k (sound waves), where c

s
is the speed
of sound (346 m s
1
here). In the dispersive case,
= (g
k)
12
(deep-
water waves). In both cases, two waves of wavenumber

k
1
= 10 m
1
and
k
2
= 40 m
1
, respectively, and amplitude A
w.1
= A
w.2
= 1, are summed
to form a group with displacement D(x. t) = A
w.1
sin(
k
1
x
.1
t) +
A
w.2
sin(
k
2
x
.2
t). Time 1 in the gures occurs when t = 0 s in both
cases, and time 2 occurs when t = 0.5 s in the nondispersive case and 4 s
in the dispersive case. In the nondispersive case, the group speed equals the
phase speed of each harmonic wave in the group and the wave group does
not change shape over time as it travels over distance x. In the dispersive
case, the group speed differs from the phase speed of each harmonic wave,
and the group changes shape over time.
For a medium to be nondispersive, the phase speed of each wave in a group
must be independent of wavenumber (or wavelength). In other words,
c
k
=
c
l
=
c
m
= 0 (4.103)
Substituting these terms into (4.101) gives the group scalar velocities in a nondis-
persive medium as
c
g.x
= c
k
|
K|
c
g.y
= c
l
|
K|
c
g.z
= c
m
|
K|
(4.104)
c
g
= c
. Thus, in a nondispersive medium,

the group speed equals the phase speed of individual waves. In the case of sound
waves originating from a stationary source, the phase speed (c
) is the speed of
sound (c
s
).
If waves traveling as a group propagate with a group speed that differs from
the phase speed of any individual wave in the group, the shape of the group
deforms over time, and the medium in which the group propagates is called a
dispersive medium. Dispersive-medium waves arise when the phase speed is a
function of wavenumber. Water waves are examples of dispersive waves (Example
4.16). Figure 4.18 shows examples of wave pulses in nondispersive and dispersive
media.
121
Example 4.16
Determine the phase speed and group speed of deep-water waves, which have
a dispersion relation,
k.
SOLUTION
Substituting the dispersion relationship into (4.97) gives the phase speed of
deep-water waves as
c
k
=
k
Substituting the dispersion relationship into (4.100) and the result into
(4.102) gives the group speed as
c
g
= c
g.x
=

k
=
1
2
k
=
1
2
c
Thus, individual deep-water waves in a group move twice as fast as the

envelope of the group.
In the following subsections, atmospheric wave types and their characteristic
dispersion relationships are identied and briey described.
4.3.5.1 Pure acoustic waves
Pure acoustic (sound) waves occur when a vibration causes alternating adiabatic
compression and expansion of a compressible uid, such as air. During compression
and expansion, air pressure oscillates, causing acceleration to oscillate along the
direction of propagation of the wave. Pure acoustic waves are longitudinal because
wave disturbances travel parallel to the direction of wave propagation. Acoustic
waves transport not only sound, but also parcels of air containing energy, gases,
and particles.
The dispersion relationship for pure acoustic waves can be derived in three
dimensions by solving the u-, v-, and w-momentum equations (ignoring the Cori-
olis, gravitational, viscous, and eddy diffusion terms), the continuity equation for
air, and the thermodynamic energy equation (ignoring diabatic energy sources and
sinks). If the direction of wave propagation is assumed to be along the x-axis, the
dispersion relationship can be simplied by setting v- and w-terms to zero. In this
case, the u-momentum, continuity, and thermodynamic energy equations simplify
from (4.73), (3.23), and (3.66) to
du
dt
=
1
a
p
a
x
(4.105)
d
a
dt
=
a
u
x
(4.106)
1
v
d
v
dt
=
dln
v
dt
= 0 (4.107)
122
respectively. Substituting
v
= T
v
(1000p
a
)
and p
a
=
a
R
T
v
into (4.107) gives
d
a
dt
=

a
dln p
a
dt
(4.108)
where = 1(1 ) = c
p.d
c
p.d
c
v.d
1.4. Substituting (4.108) into (4.106)
gives
dln p
a
dt
=
1
p
a
dp
a
dt
=
u
x
(4.109)
Taking the time derivative of (4.109), then substituting in (4.105), p
a
=
a
R
T
v
,
p
a
= p
a
+ p
a
,
a
=
a
+
a
, and u = u +u
, and eliminating products of perturba-

tion variables yield the acoustic wave equation,
d
2
p
a
dt
2
=

t
+ u

x
2
p
a
= c
2
s
2
p
a
x
2
(4.110)
where
c
s
=
T
v
(4.111)
is the adiabatic speed of sound.
Example 4.16
When T
v
= 298 K, c
s
= 346 m s
1
. When T
v
= 225 K, c
s
= 301 m s
1
.
A sinusoidal solution to (4.110) is
p
a
= p
a.0
sin(
kx
t) (4.112)
where x is the distance along the direction of phase propagation, t is time, and p
a.0
is the amplitude of the oscillation. Substituting this solution into (4.110) gives the
dispersion relationship for pure acoustic waves,
= ( u c
s
)

k (4.113)
Substituting (4.113) into (4.97) and solving give the phase speed of sound waves.
Taking the partial derivative of (4.113) with respect to

k gives the group scalar
velocity in the x-direction. The result in both cases is identical:
|c
g
| = c
g.x
= c
= u c
s
(4.114)
123
Since their group speed equals their phase speed, acoustic waves are nondispersive-
medium waves. The group speed equals the group scalar velocity in the x-direction
only because sound waves do not propagate in the y- and z-directions in this case.
4.3.5.2 Gravity, acoustic-gravity, and Lamb waves
When the atmosphere is stably stratied and a parcel of air is displaced vertically,
buoyancy restores the parcel to its equilibrium position in an oscillatory man-
ner. The frequency of oscillation, N
bv
, is the BruntV ais al a frequency, dened in
Chapter 2. The period of oscillation is
bv
= 2N
bv
. The wave resulting from this
oscillation is a gravity (buoyancy) wave. Air-parcel displacements that give rise to
gravity waves may result from forced convection, air ow over mountains, wind
shear in frontal regions, wind shear associated with jet-stream ow, or perturba-
tions associated with the geostrophic wind. Because gravity waves displace parcels
of air, they perturb temperatures, gas and particle concentrations, vertical scalar
velocities, and other parameters associated with the parcel. Gravity-wave motion
may increase vertical transfer of ozone between the lower stratosphere and upper
troposphere (e.g., Lindzen 1981; van Zandt and Fritts 1989). Gravity waves may
also cause large periodic uctuations in ozone at specic locations in the tropo-
sphere (Langford et al. 1996).
The dispersion relationship for gravity waves can be isolated froma broader rela-
tionship for acoustic-gravity waves. In three dimensions, the acoustic-gravity-wave
dispersion relationship is found by solving the three momentum equations (retain-
ing gravity in the vertical), the continuity equation for air, and the thermodynamic
energy equation. If only the xz plane is considered, a simplied acoustic-gravity
wave relationship is obtained by solving
du
dt
=
1
a
p
a
x
dw
dt
=
1
a
p
a
z
g
d
a
dt
=
a
u
x
+
w
z
d
a
dt
=

a
dln p
a
dt
(4.115)
written from (4.73), (4.75), (3.23), and (4.108), respectively. When p
a
= p
a
+ p
a
,
a
=
a
+
a
, u = u +u
, and w = w+w
, (4.115) can be solved analytically to

obtain the acoustic-gravity wave dispersion relationship,
N
2
bv
(
k)
2
k
2
+
(
k)
2
c
2
s
= m
2
+

k
2
+

2
c
c
2
s
(4.116)
where
c
=
c
s
2H
(4.117)
124
is the acoustic cutoff frequency (s
1
). In this equation, H is the scale height of the
atmosphere, obtained from (2.47) under the assumption that the atmosphere is
isothermal and stably stratied. The acoustic cutoff frequency is a frequency above
which acoustic waves may propagate vertically if other conditions are right. Below
this frequency, acoustic waves cannot propagate vertically.
Example 4.17
For an isothermal layer at T
v
= 298 K, nd the acoustic cutoff and Brunt
V ais al a frequencies.
SOLUTION
From Appendix A,

M = 4.8096 10
26
kg, g = 9.81 ms
2
, and k
B
= 1.3807
10
23
kg m
2
s
2
K
1
molec.
1
. From (2.47), the scale height is H = 8.72 km.
Thus, from (4.111) c
s
= 346 m s
1
and from (4.117),
c
= 0.0198 s
1
. From
(2.106), the BruntV ais al a frequency is N
bv
=
g (I
d
I
v
)T
v
. Substituting
I
v
= 0 (since the atmosphere is isothermal) and I
d
= +9.8 K km
1
into this
equation gives N
bv
= 0.0180 s
1
. Thus, the acoustic cutoff frequency slightly
exceeds the BruntV ais al a frequency (
c
> N
bv
) in this case.
Dispersion relationships for gravity waves, acoustic waves modied by the effects
of stable stratication, and mixed waves can be isolated from the acoustic-gravity-
wave dispersion relationship by considering certain frequency and wavenumber
regimes. For example, (4.116) simplies for several cases to
k +
N
bv
k
2
+ m
2
+
2
c
c
2
s
12

2

2
c
low-frequency
gravity waves
u
k +
c
2
s

k
2
+c
2
s
m
2
+
2
c
12
N
2
bv
high-frequency
acoustic waves
u
k +
N
bv
k
(
k
2
+ m
2
)
12
k {mountain lee waves

( u +c
s
)

2
c
. m
2
= 0.

k 0
N
2
bv
. m
2
= 0.

k
{ Lamb waves
(4.118)
When
2

2
c
, (4.116) simplies to the dispersion relationship for low-
frequency gravity waves. The same result is found by assuming the atmosphere
is incompressible (c
s
). Figure 4.19 shows that vertically and horizontally
propagating, low-frequency gravity waves occur when m
2
> 0,
k - c
s
k as
k 0, and
k - N
bv
as

k . Such waves are internal gravity waves. An
internal wave is a wave that exhibits vertical motion trapped beneath a surface. The
125
0
0.01
0.02
0.03
0.04
0.05
0.06
0.07
0.08
0 5 10
5
0.0001 0.00015 0.0002
v

(
s
1
)
k
2
= 0
~
m
2
> 0
~
m
2
< 0
~
m
2
= 0
~
m
v
c
N
bv
Lamb waves
High-frequency internal
acoustic waves
Low-frequency internal
gravity waves
~
2
> 0
~
m
(m
1
)
u
k
~
c k
s
~
Figure 4.19 Vertical wavenumber squared, m
2
, as a func-
tion of

k and
when H = 8.72 km, c

s
= 346 m s
1
, and
N
bv
= 0.0180 s
1
(see Example 4.17). The upper region
where m
2
> 0 represents conditions that give rise to ver-
tically propagating high-frequency acoustic waves. The
lower region where m
2
> 0 represents conditions that give
rise to vertically propagating low-frequency gravity waves.
The region in which m
2
- 0 represents conditions that
give rise to horizontally propagating acoustic-gravity and
Lamb waves. Lamb waves connect the point

k = 0, m
2
=0
to the point

k = , m
2
= 0. The acoustic cutoff frequency
slightly exceeds the BruntV ais al a frequency.
BruntV ais al a frequency demarks the upper frequency for internal gravity waves.
Equating (4.97) with the rst termon the right side of (4.118) gives the phase speed
of internal gravity waves as
c
= u +
N
bv
k
2
+ m
2
+
2
c
c
2
s
(4.119)
Applying (4.100) to this phase speed gives group scalar velocities of internal gravity
waves as
c
g.x
= u +
N
bv
m
2
+
2
c
c
2
s
k
2
+ m
2
+
2
c
c
2
s
32
c
g.z
=
N
bv
k m
k
2
+ m
2
+
2
c
c
2
s
32
(4.120)
Thus, internal gravity waves are dispersive-medium waves, since their phase speed
differs from the magnitude of their group velocity vector.
In the region of Fig. 4.19 where m
2
- 0, waves are external acoustic-gravity
waves and propagate only horizontally. An external wave is a wave that has its
maximum amplitude at the external boundary (free surface) of a uid. Such waves
have no inuence in the vertical, since their amplitudes decrease exponentially with
distance from the surface.
126
When
2
N
2
bv
, the frequency from(4.118) simplies to that of acoustic waves,
modied by the effects of stable stratication. Figure 4.19 shows that vertically
and horizontally propagating acoustic waves occur when m
2
> 0,
k >
c
as

k 0, and
k > c
s
k as

k . Such waves are high-frequency internal
acoustic waves. The dispersion relationship for internal acoustic waves differs from
that for pure acoustic waves only in that the latter ignores the effect of buoyancy
in a stably-stratied atmosphere.
The dotted line in Fig. 4.19 connecting the points between m
2
= 0,

k 0 for
internal gravity waves and m
2
= 0,

k for internal acoustic waves modied
by stratication describes the dispersion relationship for external Lamb waves
(Lamb 1910). The frequency of a Lamb wave approaches
( u +c
s
)

k, which
is the same frequency as that of a pure acoustic wave, as shown in (4.113). Lamb
waves are nondispersive and propagate horizontally at the speed of sound. Whereas
their energy decreases exponentially with increasing altitude, their amplitude may
decrease or increase exponentially with altitude, depending on the value of m.
Mountain lee waves are gravity waves formed as a wind is perturbed vertically
as it ows over a mountain in a stable atmosphere. Such waves have a wavelength
on the order of 10 km, which results in a wavenumber of

k 6.3 10
4
m
1
. This
is large in comparison with
c
c
s
5.7 10
5
m
1
, obtained from Example 4.17.
Thus, in the case of mountain lee waves,

k
c
c
s
, and the acoustic-gravity-wave
dispersion relationship simplies from that of low-frequency gravity waves to that
of mountain lee waves, as shown in (4.118).
4.3.5.3 Inertia gravity waves and inertia Lamb waves
After pressure and buoyancy oscillations, a third type of oscillation is an inertial
oscillation. When a parcel of air moving from west to east with the geostrophic
wind is perturbed in the southnorth direction, the Coriolis force propels the parcel
toward its original latitude in an inertially stable atmosphere and away from its
original latitude in an inertially unstable atmosphere. In the former case, the parcel
subsequently oscillates about its initial latitude in an inertial oscillation.
Whether the atmosphere is inertially stable or unstable can be estimated by
solving the horizontal momentum equations,
du
dt
= f v = f
dy
dt
(4.121)
dv
dt
= f (u
g
u) (4.122)
where only acceleration, the Coriolis force, and the southnorth pressure gradient
(embodied in u
g
) are considered. If a parcel of air moving with the geostrophic wind
is displaced in the southnorth direction a distance Ly, (4.121) can be integrated
between y
0
and y
0
+Ly to give
u(y
0
+Ly) u
g
(y
0
) f Ly (4.123)
127
The geostrophic wind at point y
0
can be found by taking a rst-order Taylor series
expansion of the geostrophic wind at y
0
+Ly, giving
u
g
(y
0
+Ly) u
g
(y
0
) +
u
g
y
Ly (4.124)
Substituting (4.124) into (4.123) yields
u
g
(y
0
+Ly) u(y
0
+Ly)
f
u
g
y
Ly (4.125)
which substitutes into (4.122) to give
dv
dt
= f
f
u
g
y
Ly (4.126)
This equation states that, if f u
g
y > 0 and a parcel of air is displaced north-
ward of its original latitude in the Northern Hemisphere (Ly > 0), then dvdt - 0,
and the parcel decelerates, turns around, and begins to accelerate southward. As the
parcel passes south of its original latitude (Ly - 0), then dvdt > 0, and the parcel
decelerates, turns around, and accelerates northward again, and so on. Thus, when
f u
g
y > 0 a southnorth perturbation to an air parcel causes it to oscillate
inertially about its original latitude. Under such circumstances, the atmosphere is
inertially stable.
If f u
g
y - 0 and an air parcel is displaced northward (Ly > 0), then
dvdt > 0, and the parcel continues to accelerate northward. Similarly, if it is dis-
placed southward, dvdt - 0, and it continues to accelerate southward. Under
these conditions, the atmosphere is inertially unstable.
Finally, when f u
g
y = 0, a perturbation to a parcel has no effect on south
north acceleration (dvdt = 0). The atmosphere is inertially neutral in this scenario.
In sum, inertial stability criteria for the Northern Hemisphere, where f > 0, can
be dened with
f
u
g
y
- 0 inertially unstable
= 0 inertially neutral
>0 inertially stable
(4.127)
which are analogous to the buoyancy-related stability criteria discussed in
Chapter 2.
The atmosphere is almost always inertially stable, except near the Equator, where
the Coriolis parameter is small. When inertial oscillations occur, their frequency f
is on the order of 10
4
s
1
. This is much lower than the acoustic cutoff fre-
quency (
c
) and BruntV ais al a frequency (N
bv
), which have values on the order of
10
2
s
1
(Example 4.17). Thus, inertial oscillations are important only for pertur-
bations involving low-frequency (long-wavelength) waves.
128
Since vertically propagating acoustic waves occur only at high frequency (>
c
,
as shown in Fig. 4.19), they are not affected much by inertial oscillations. When
the frequency of a Lamb wave is below
c
, it may be affected by inertial oscilla-
tions. Under such a condition, the dispersion relationship for inertia Lamb waves
is rewritten from (4.118) to
= f
2
+c
2
s

k
2
(4.128)
Similarly, the dispersion relationship for inertia gravity waves is rewritten from
that of low-frequency gravity waves in (4.118) to
= f
2
+
N
2
bv
k
2
k
2
+ m
2
+
2
c
c
2
s
(4.129)
Inertia gravity waves are driven by inertial and buoyant oscillations and occur
when the atmosphere is inertially and buoyantly stable. Inertia Lamb waves are
driven by inertial and pressure oscillations.
Inertia Lamb waves and inertia gravity waves participate in the restoration of
the geostrophic wind following a perturbation to it. When the geostrophic wind
in midlatitudes, which ows from west to east, is perturbed to the north or south,
horizontally propagating inertia Lamb waves act to restore vertically integrated
geostrophic velocity and surface pressure elds while vertically and horizontally
propagating inertia gravity waves act to restore vertical shear and temperature elds
to geostrophic balance (Arakawa 1997). The restoration to geostrophic balance is
called geostrophic adjustment.
During geostrophic adjustment, the primary factor that determines whether the
pressure (mass) eld adjusts to the velocity eld or vice-versa is the size of the length
scale of horizontal motion (L) relative to the Rossby radius of deformation. For an
atmospheric wave, the length scale is its horizontal wavelength. The Rossby radius
of deformation is
R
=
gh
e
f
(4.130)
where
h
e
=
c
2
s
g
inertia Lamb waves
N
2
bv
g
m
2
+
2
c
c
2
s
inertia gravity waves
(4.131)
is the equivalent depth of the atmosphere. A typical value of
R
for inertia Lamb
waves is 3000 km, and typical values for inertia gravity waves are 1001000 km.
129
H
L
Geostrophic
wind
570 hPa 566 hPa
East
x
y
Pressure-
gradient
force
Coriolis
force
150 km
Figure 4.20 Geostrophic adjustment of
the velocity eld to the pressure eld at
one altitude when L >
R
and when the
domain is assumed to be large enough
to allow the energy of the oscillations
to disperse and decay. In reality, the ver-
tical mean velocity eld, not the veloc-
ity eld at a given altitude, adjusts to
the surface pressure. A perturbation to
the geostrophic wind causes waves to
restore the system to geostrophic bal-
ance along an oscillatory path.
Example 4.18
From the information in Example 4.17, calculate the equivalent depth and the
Rossby radius of deformation for inertia gravity waves assuming
.z
= 5 km
and = 30
N latitude.
SOLUTION
From Example 4.17, c
s
= 346 m s
1
,
c
= 0.0198 s
1
, and N
bv
= 0.0180 s
1
,
since m= 2
.z
= 0.0013 m
1
. From (4.131), h
e
= 19.5 m. At = 30
N, we
have f = 7.292 10
5
s
1
and
R
= 190 km.
When inertia Lamb waves are present and L >
R
, the perturbed vertical mean
velocity eld adjusts to the pressure eld, as roughly demonstrated in two dimen-
sions in Fig. 4.20. When L -
R
, the pressure eld adjusts to the velocity eld. When
L =
R
, a mutual adjustment between the pressure and velocity elds occurs.
When inertia gravity waves are present and L >
R
, the wind shear eld adjusts
to the temperature eld. Otherwise the temperature eld adjusts to the wind shear
eld or mutual adjustment occurs (Arakawa 1997).
130
4.3.5.4 Rossby waves
Rossby (planetary) waves (Rossby et al. 1939) arise from a change in the Coriolis
parameter f = 2Osin with latitude. These waves are important for synoptic and
planetary-scale meteorology. If an air parcel moving from west to east is perturbed
to the south or north, the change in f with latitude provides a restoring force that
gives rise to freely propagating Rossby waves. The presence of topography can
initiate forced topographic Rossby waves. Rossby waves differ from pure inertial
oscillations in that pure inertial oscillations do not require a variation of f with
latitude.
In an incompressible, frictionless, adiabatic atmosphere of constant depth, abso-
lute vorticity of air is conserved. Rossby waves can be described in terms of the
conservation of absolute vorticity. Vorticity is the measure of spin around an axis.
The faster a body spins, the greater its vorticity. Air spins relative to the surface of
the Earth, and the surface of the Earth spins relative to its own axis of rotation. The
spin of air relative to the surface of the Earth is relative vorticity, and the spin of the
Earth relative to its own axis is the Earths vorticity. Positive vorticity corresponds
to counterclockwise spin when the spin is viewed from above the North Pole. Since
the Earth rotates counterclockwise, the Earths vorticity is positive. For a body at
the North Pole, the Earths vorticity is maximum, since the body spins around its
own vertical axis. For a body standing at the Equator, the Earths vorticity is zero,
since the body does not spin around its own vertical axis. The vertical component
of the Earths vorticity is the Coriolis parameter f. At the Equator, North Pole, and
South Pole, f = 0, 2O, and 2O, respectively.
The relative vorticity of air is the vorticity of air relative to the Earth. Mathe-
matically, the relative vorticity is the curl of velocity. The relative vorticity vector
is
r
= v =
i j k
z
u v w
w
y

v
z
w
x

u
z
j +
v
x

u
y
k
(4.132)
The absolute vorticity is the sum of the Earths and the relative vorticity. The
vertical component of absolute vorticity is
a.z
= f +
r.z
, where
r.z
is the vertical
component of relative vorticity. In a frictionless, incompressible, and adiabatic
atmosphere, the vertical component of absolute vorticity divided by the depth of
the atmosphere is constant. In other words,
P
v
=
f +
r.z
Lz
t
=
f +
v
x

u
y
Lz
t
= constant (4.133)
where P
v
is a simplied form of potential vorticity and Lz
t
is the depth of the
atmosphere from the surface to the tropopause. If the depth of the atmosphere is
constant and v is nonzero, a freely propagating Rossby wave can develop as follows:
As air moves north, f increases. To conserve potential vorticity,
r.z
decreases. For
131
r.z
to decrease, u and v must change in a way that forces the particle south, and
so on. The sinusoidal motion resulting from the continuous changes of u, v, and f
is a freely propagating Rossby wave.
The dispersion relationship for freely propagating Rossby waves in a frictionless,
incompressible, and adiabatic atmosphere is found by solving the u and v momen-
tum equations and the continuity equation for air. For this derivation, atmospheric
depth is allowed to vary. The constant-depth result is easily extracted from the
solution.
If only inertial, Coriolis, and pressure-gradient terms are considered, the hori-
zontal momentum equations simplify from (4.73) and (4.74) to
du
dt
= f v
1
p
a
x
z
(4.134)
dv
dt
= f u
1
p
a
y
z
(4.135)
where the subscript z indicates the partial derivative is taken along a surface of
constant altitude. At least three sets of terms can be substituted into these equations.
The rst is the midlatitude beta-plane approximation,
f = f
0
+(y y
0
) (4.136)
which is an approximation for the variation of the Coriolis parameter with latitude.
In this equation, y
0
is the southnorth distance from the Equator at which f = f
0
,
y is the southnorth distance of interest from the Equator, (y y
0
) f
0
,
and
=
f
y
= 2O
y
cos
2O
R
e
cos (4.137)
where y
= R
e
.
The second is a pair of equations relating pressure gradients along surfaces of
constant altitude to geopotential gradients along surfaces of constant pressure. The
equations, derived in Section 5.3, are
p
a
x
z
=
a
+
x
p
(4.138)
p
a
y
z
=
a
+
y
p
(4.139)
where p indicates that a partial derivative is taken along a surface of constant
pressure.
The third is a set of equations dividing scalars (u, v, +) into geostrophic (u
g
,
v
g
, +
g
) and ageostrophic (u
a
, v
a
, +
a
) components. Substituting (4.138), (4.136),
u = u
g
+u
a
, and + = +
g
++
a
into (4.134) gives
d(u
g
+u
a
)
dt
= [ f
0
+(y y
0
)] (v
g
+v
a
)
(+
g
++
a
)
x
p
(4.140)
132
Substituting (4.139), (4.136), v = v
g
+v
a
, and + = +
g
++
a
into (4.135) gives
d(v
g
+v
a
)
dt
= [ f
0
+(y y
0
)] (u
g
+u
a
)
(+
g
++
a
)
y
p
(4.141)
These equations can be simplied by combining (4.138) and (4.139) with (4.79) at
latitude y
0
to give geostrophic scalar velocities along surfaces of constant pressure
as
v
g
=
1
f
0
+
g
x
p
u
g
=
1
f
0
+
g
y
p
(4.142)
Substituting these terms into (4.140) and (4.141) and assuming du
a
dt, dv
a
dt,
(y y
0
) u
a
, and (y y
0
) v
a
are small give the quasigeostrophic horizontal
momentum equations,
du
g
dt
= f
0
v
a
+(y y
0
) v
g
+
a
x
p
(4.143)
dv
g
dt
= f
0
u
a
(y y
0
) u
g
+
a
y
p
(4.144)
Subtracting y of (4.143) from x of (4.144) yields
d
dt
v
g
x

u
g
y
= f
0
u
a
x
+
v
a
y
v
g
(4.145)
where all remaining terms cancel. For example, u
g
x +v
g
y = 0, which indi-
cates that the geostrophic wind is nondivergent.
The vertical scalar velocity in the altitude coordinate is
w =
dz
dt
=
1
g
d+
dt
(4.146)
Substituting this expression, u = u
g
+u
a
, v = v
g
+v
a
, and u
g
x +v
g
y = 0
into the incompressible continuity equation from (3.27), and assuming d+
a
dt is
small give
1
g
d+
g
dt
u
a
x
+
v
a
y
(4.147)
Integrating (4.147) from the surface (z = 0) to the mean tropopause height,
z = Lz
t
, gives
d+
g
dt
= gLz
t
u
a
x
+
v
a
y
(4.148)
When d+
g
dt = 0, the atmosphere has a free surface and average depth Lz
t
. When
d+
g
dt = 0, the atmosphere has a rigid lid with constant depth Lz
t
. Substituting
133
(4.148) into (4.145) gives
d
dt
f
0
gLz
t
+
g
= v
g
(4.149)
where
g
=
v
g
x

u
g
y
=
1
f
0
2
+
g
x
2
+

2
+
g
y
2
p
=
2
p
+
g
f
0
(4.150)
is the geostrophic potential vorticity.
Substituting (4.150) and v
g
from (4.142) into (4.149) and expanding the total
derivative give the quasigeostrophic potential vorticity equation,

t
+ u

x
2
p
+
g
f
2
0
gLz
t
+
g
+
g
x
p
= 0 (4.151)
where u and v were replaced by u = u +u
and v v
, respectively, and the equa-

tion was Reynolds-averaged. The assumption v v
is reasonable for midlatitudes,

where u dominates over v. Equation (4.151) has a sinusoidal wave solution,
+
g
= +
g.0
sin
kx +
l y
(4.152)
where +
g.0
is the amplitude of the wave. Substituting (4.152) into (4.151) and
solving for
give the dispersion relationship for freely propagating Rossby waves

in a frictionless, incompressible, and adiabatic atmosphere as
k
2
+
l
2
+
2
R
k (4.153)
where
R
=
gLz
t
f
0
is another form of the Rossby radius of deformation.
For an atmosphere of constant depth, d+
g
dt = 0 in (4.149), and (4.153)
simplies to
k
2
+
l
2
k (4.154)
Both dispersion relationships suggest that Rossby-wave phase speeds, which are less
than zero in the equations, propagate from east to west (Example 4.19). However,
since the observed mean zonal wind u is fromwest to east with a magnitude greater
than the Rossby-wave phase speed, Rossby waves plus the mean wind travel from
west to east.
134
4.4 Summary
Example 4.19
From (4.154), estimate the Rossby-wave phase speed when
.z

.y
= 6000
km at = 45
N.
SOLUTION
From(4.95), k l = 2
.z
=1.05 10
6
m
1
. From(4.137), =1.62 10
11
m
1
s
1
. From (4.154), c
k
2
+
l
2
= 7.3 m s
1
. Thus, Rossby-wave
phase speed propagates from east to west.
The wave solution for Rossby waves accounts for horizontally propagating
waves in an incompressible atmosphere. Rossby waves propagate vertically as well
as horizontally. To account for vertical propagation and air compressibility, the dis-
persion relationship for Rossby waves can be found by using the divergent continu-
ity equation for air (in three dimensions) and the thermodynamic energy equation
from (4.115), instead of the nondivergent continuity equation from (3.27). In the
new case, the dispersion relationship is the same as (4.153), except that the Rossby
radius of deformation
R
=
gh
e
f
0
, where the equivalent depth is the same as
that for inertia gravity waves, given in (4.131). The resulting dispersion relation-
ship for vertically and horizontally propagating Rossby waves in a compressible
atmosphere is
= u
k
2
+
l
2
+
f
2
0
N
2
bv
m
2
+

2
c
c
2
s
(4.155)
For Rossby waves that propagate only horizontally in a compressible atmosphere,
the dispersion relationship is given by (4.153) with
R
=
gh
e
f
0
, where h
e
=
c
2
s
g.
Rossby waves can be initiated, not only by a perturbation to the geostrophic
velocity in the southnorth direction, but also by airow over large topographical
barriers, such as the Rocky or Himalayan mountain ranges. As seen from (4.133),
a change in the depth of the atmosphere forces f, u, and v to change, initiating
forced topographic Rossby waves.
4.4 SUMMARY
In this chapter, elements of the momentum equation, including the local accelera-
tion, the Coriolis acceleration, the Earths centripetal acceleration, the gravitational
force, the pressure-gradient force, the viscous force, and the turbulent-ux diver-
gence, were derived. For a coordinate system xed on the surface of the Earth,
the Coriolis and the Earths centripetal acceleration are treated as apparent forces.
135
The momentum equation was simplied to derive expressions for the geostrophic
wind and surface winds. It was also transformed to cylindrical coordinates to
derive an expression for the gradient wind. Atmospheric waves were also discussed.
Important waves include acoustic, Lamb, gravity, inertia Lamb, inertia gravity, and
Rossby waves. In the next chapter, the equations of atmospheric dynamics are con-
verted to pressure, sigma-pressure, and sigma-altitude vertical coordinates.
4.5 PROBLEMS
4.1 Assume a spherical coordinate gridcell, centeredat =60
S, has dimension
d
e
= 2.5
and d = 2
. Calculate dx at the north and south boundaries of

the cell, and calculate dy.
4.2 If u = 10 m s
1
, v = 10 m s
1
, w = 0.01 m s
1
, and dudt scales as u(Lxu),
estimate each term on the right side of (4.29) at = 30
N. Let Lx = 5 km,
Ly = 5 km, and Lz = 1 km.
4.3 In Cartesian coordinates, show that (v
) v =
v
2
2
+( v) v.
4.4 Assume a grid cell has dimension Lx = 5 km, Ly = 4 km, and Lz = 0.1 km,
and assume the west, east, south, north, lower, and upper boundary scalar
velocities are u
1
= +2, u
2
= +3, v
3
= +1, v
4
= 3, w
5
= +0.03, and w
6
=
+0.04 m s
1
, respectively. Estimate the magnitude of the divergence term of
the local acceleration (v
v).
4.5 If u =30 ms
1
, v =10 ms
1
, and =45
N, nd the Coriolis-force magnitude

per unit mass.
4.6 Assume a grid cell has dimension Lx = 500 km and Ly = 400 km, centered
at = 30
N latitude. Assume potential virtual temperature and air pressure

at the west, east, south, and north boundaries are
v.1
= 302 K and p
a.1
=
520 hPa,
v.2
= 304 K and p
a.2
= 530 hPa,
v.3
= 302 K and p
a.3
= 500 hPa,
and
v.4
= 304 K and p
a.4
= 540 hPa, respectively. Estimate the geostrophic
scalar velocities and geostrophic wind speed.
4.7 In a region to the east of the center of an intense hurricane at = 20
N
latitude, a westeast pressure gradient of 50 hPa per 125 km and a gradient
wind scalar velocity of v = 70 m s
1
are observed. What is the distance
between the center of the hurricane and the observation? Assume T
v
=280 K
and p
a
= 930 hPa.
4.8 (a) Assume u
1
= 1 m s
1
at x
1
= 0 m, u
2
= 2 m s
1
at x
2
= 2500 m, u
3
=
3 m s
1
at x
3
= 5000 m, and K
m,xx
= 2.5 10
3
m
2
s
1
. If density and
K
m,xx
are constant, estimate the change in u, due to eddy diffusion alone,
at the cell center after one hour.
(b) Assume n
1
= 0.02 m s
1
at location z
1
= 0 m,
n
2
= 0.02 m s
1
at z
2
= 50 m, n
3
= 0.04 m s
1
at z
3
= 1000 m,
and K
m,zz
= 50 m
2
s
1
. If density and K
m,zz
are constant, estimate the
change in w at the cell center after 100 s, due to eddy diffusion alone.
4.9 Calculate the viscous acceleration force per unit mass when the air is dry,
T = 298 K, p
a
= 995 hPa, and u = 1.0 m s
1
at 10 cm above the ground and
0.7 m s
1
at 5 cm above the ground.
4.10 Derive (4.155).
136
4.11 (a) Assume Lx = 5 km for a grid cell, : = 5 m s
1
at the center of the cell,
= 35
N, T = 285 K, uy = 0, uz = 0, p
a.1
= 1010 hPa and p
a.2
=
1006 hPa at the west and east boundaries of the cell, and u
1
= +3 m s
1
andu
2
=+4 ms
1
at the west andeast boundaries of the cell, respectively.
Write a computer script to nd u over time from (4.73), assuming that
viscous and diffusive terms are zero. Assume ux in the uux term
is constant throughout the simulation, but u is not. Set the time step to
1 s and run the program for one hour. Plot the results.
(b) Change p
a.2
to (i) 1009 hPa and (ii) 1011 hPa. Plot the results in both
cases. For p
a.2
= 1009 hPa, change u
1
to +3.5 m s
1
and plot the results.
Comment on the sensitivity of the solution to changes in pressure and
velocity.
4.12 Write a computer programto replicate the results inFig. 4.18 for bothnondis-
persive and dispersive waves and plot the results.
4.13 Write a computer program to replicate the results in Fig. 4.19 and plot the
results.
137
5
Vertical-coordinate conversions
T
he equations of atmospheric dynamics can be solved either in fully com-
pressible form or after they are simplied with the hydrostatic, anelastic, or
Boussinesq approximation. In many cases, the equations are adapted to a vertical
coordinate other than the altitude coordinate. Such coordinates include the pres-
sure, sigma-pressure, sigma-altitude, and isentropic coordinates. The rst three are
discussed in this chapter. A disadvantage of the altitude and pressure coordinates is
that both allow model layers to intersect ground topography. The sigma-pressure
and sigma-altitude coordinates are terrain-following and do not allowmodel layers
to intersect ground topography. A disadvantage of the pressure and sigma-pressure
coordinates is that both assume the atmosphere is in hydrostatic balance. Such an
assumption is reasonable for global and most mesoscale models but causes inaccu-
racy for ne horizontal grid resolution, where the atmosphere is often nonhydro-
static. For nonhydrostatic simulations, the altitude or sigma-altitude coordinate
can be used.
5.1 HYDROSTATIC AND NONHYDROSTATIC MODELS
Explicit numerical schemes require a time step smaller than the minimum grid
spacing divided by the speed of the fastest motion in the grid domain. When acoustic
waves are present, they are the fastest motion in the domain, and the time step
required to resolve these waves is generally limited by the vertical grid spacing
(which is usually smaller than the horizontal grid spacing) divided by the speed
of sound. As shown in (4.118), vertically propagating acoustic waves arise as an
analytical solution to the equations of atmospheric dynamics when the equations
include the horizontal and vertical momentum equations, the continuity equation
for air, and the thermodynamic energy equation.
To eliminate the need for a short time step, vertically propagating acoustic waves
must be removed (ltered out) as a solution to the basic equations. This can be
accomplished by removing the total derivative (dwdt) fromthe vertical momentum
equation or removing the local derivative (
a
t) fromthe continuity equation for
air. Removing dwdt from the vertical momentum equation gives the hydrostatic
approximation. Removing
a
t from the continuity equation for air gives the
anelastic approximation (e.g., Ogura and Phillips 1962; Pielke 1984; Bannon
1996). A subset of the anelastic approximation, used for shallow ows only, is
the Boussinesq approximation.
138
5.1 Hydrostatic and nonhydrostatic models
Amodel that uses the hydrostatic equation instead of the full vertical momentum
equation is a hydrostatic model. The hydrostatic approximation lters out vertically
propagating acoustic waves in a hydrostatic model. A model that retains the full
vertical momentum equation is a nonhydrostatic model. A nonhydrostatic model
may be anelastic (it solves the anelastic continuity equation for air) or elastic (solves
the full continuity equation for air and is fully compressible). The anelastic and
Boussinesq approximations lter out vertically propagating acoustic waves in an
anelastic nonhydrostatic model. Vertically propagating acoustic waves are retained
in an elastic nonhydrostatic model.
In a hydrostatic model and in an anelastic or Boussinesq nonhydrostatic model,
horizontally propagating acoustic waves still exist but vertically propagating acous-
tic waves do not. Thus, when either model is solved explicitly, its time step is lim-
ited by the smaller of the minimum horizontal grid spacing divided by the speed
of sound and the minimum vertical grid spacing divided by the maximum vertical
scalar velocity. In most cases, the smaller value is the minimum horizontal grid
spacing divided by the speed of sound.
Example 5.1
When T
v
= 298 K, the maximum vertical scalar velocity is 10 cm s
1
, the
minimum vertical grid spacing is 10 m, and the minimum horizontal grid
spacing is (a) 500 km, (b) 5 km, and (c) 5 m, nd the maximum time step in a
(i) hydrostatic model and (ii) elastic nonhydrostatic model when both models
are solved explicitly.
SOLUTION
From (4.111), the adiabatic speed of sound at 298 K is c
s
= 346 m s
1
.
(i) For the hydrostatic model, the time step is the smaller of 100 s (the vertical
dimension divided by the vertical scalar velocity) and (a) 1450 s, (b) 14.5 s,
or (c) 0.014 s (the horizontal grid spacing divided by the speed of sound),
giving the time step in each case as (a) 100 s, (b) 14.5 s, or (c) 0.014 s.
(ii) For the nonhydrostatic model, the time step is the smaller of 0.028 s (the
vertical dimension divided by the speed of sound) and (a) 1450 s, (b) 14.5 s,
or (c) 0.014 s (the horizontal dimension divided by the speed of sound),
giving the time step in each case as (a) 0.028 s, (b) 0.028 s, and (c) 0.014 s.
Thus, the time step required in an elastic nonhydrostatic model solved explic-
itly is much smaller than is that in a hydrostatic model unless the horizontal
grid spacing is equal to or less than the vertical grid spacing in both models.
In an elastic nonhydrostatic model solved explicitly, the time step is limited by
the minimum horizontal or vertical grid spacing divided by the speed of sound. In
most mesoscale models, the minimumhorizontal grid spacing is 1050 times greater
than the vertical grid spacing, so the time step in an explicit elastic nonhydrostatic
139
model is proportionately smaller than in a hydrostatic or elastic model. Below,
the equations required in a hydrostatic, an elastic nonhydrostatic, an anelastic
nonhydrostatic, and a Boussinesq nonhydrostatic model are discussed briey.
5.1.1 Hydrostatic model
A hydrostatic model is a model in which dwdt = 0 in the vertical momentum
equation. Such a model requires the solution of the following equations. The poten-
tial virtual temperature is found from the thermodynamic energy equation given in
(3.76). The water vapor specic humidity is found fromthe species continuity equa-
tion given in (3.61). Horizontal scalar velocities are determined from the momen-
tum equations given in (4.73)(4.74) and (4.76)(4.77). Air pressure is extracted
diagnostically from the hydrostatic equation given in (2.40). Air temperature is
calculated diagnostically from potential virtual temperature with (2.96). Density is
found as a function of pressure and temperature from the equation of state given
in (2.36). Once density is known, w is extracted diagnostically from the continuity
equation for air, given in (3.55). An advantage of the hydrostatic approximation is
that it removes vertically propagating acoustic waves as a possible solution to the
equations of atmospheric dynamics. A disadvantage is that it does not resolve ver-
tical accelerations well. Such accelerations become more important as grid spacing
becomes ner.
5.1.2 Elastic nonhydrostatic model
An elastic nonhydrostatic model is a model that solves the fully compressible equa-
tions of atmospheric dynamics without the anelastic or hydrostatic assumption.
The equations solved are the same as for the hydrostatic model, except that the
vertical scalar velocity is found from the vertical momentum equation, (4.75) or
(4.78), air density is found from the continuity equation for air, and air pressure is
found from the equation of state.
When an explicit (as opposed to implicit or semiimplicit Chapter 6) time-
differencing scheme is used to solve elastic equations, a short time step is required
to resolve acoustic-wave perturbations. A method to reduce computational time in
such cases is to solve terms producing acoustic waves with a shorter time step than
other terms (e.g., Klemp and Wilhelmson 1978; Chen 1991; Dudhia 1993; Wicker
and Skamarock 1998, 2002). Alternatively, an implicit or semiimplicit scheme can
be used to integrate terms responsible for acoustic waves implicitly (e.g., Tapp and
White 1976; Tanguay et al. 1990; Golding 1992; Skamarock and Klemp 1992).
5.1.3 Anelastic nonhydrostatic model
An anelastic nonhydrostatic model is a model in which
a
t =0 in the continuity
equation for air. The potential virtual temperature, specic humidity, horizontal
scalar velocities, temperature, and density are found from the same equations as in
140
5.1 Hydrostatic and nonhydrostatic models
a hydrostatic model, but pressure and the vertical scalar velocity are derived from
new equations.
The pressure is determined from an equation for nonhydrostatic pressure,
derived by assuming that the pressure at a given location can be decomposed as
p
a
= p
a
+ p
a
(5.1)
In this equation, p
a
is an average pressure, integrated horizontally over a large-scale,
hydrostatic environment, and p
a
( p
a
) is the difference between the actual pres-
sure and the large-scale pressure. In a hydrostatic model, p
a
= 0. Other variables,
including density (
a
), specic volume (
a
= 1
a
), and potential virtual tempera-
ture (
v
), are decomposed in a similar manner. In the large-scale environment, the
atmosphere is assumed to be in hydrostatic balance. Thus, the following equality
holds:
1

a
p
a
z
=
a
p
a
z
= g (5.2)
At any given location, the air is not in hydrostatic balance, and the full vertical
momentumequation must be considered. Decomposing gravitational and pressure-
gradient terms in the vertical momentum equation of (4.75) and substituting (5.2)
into the result give
g +
1
a
p
a
z
= g +(
a
+
a
)

z
( p
a
+ p
a
)
a
p
a
z

a

a
g (5.3)
where the product of perturbation terms has been removed because it is small.
Substituting (5.3) into (4.75) and ignoring eddy diffusion for now give
w
t
= u
w
x
v
w
y
w
w
z

a
p
a
z
+

a

a
g (5.4)
Taking the divergence of the sumof (5.4) and the horizontal momentumequations,
(4.73) and (4.74), results in
(v
a
) =
[
a
(v
) v]
(
a
f k v)
2
z
p
a
2
p
a
+ g

z
a

2
a
(5.5)
where and
z
are total and horizontal gradient operators from (3.6) and (4.81),
respectively. Equation (5.5) can be simplied by differentiating p
a
a
= R
T
v
to
obtain d
a
a
= dT
v
T
v
dp
a
p
a
, substituting T
v
=
v
( p
a
1000 hPa)
from(2.96)
into the result, and replacing differential and nondifferential terms with perturba-
tion and large-scale terms, respectively. These steps yield
a

a
c
v.d
c
p.d
p
a
p
a
(5.6)
Another simplication to (5.5) is obtained by removing the local derivative
in the continuity equation for air and assuming that density in the equation is
141
large-scale density. The result is the anelastic approximation to the continuity
equation,
(v
a
) = 0 (5.7)
Substituting (5.6) and (5.7) into (5.5), adding the eddy diffusion term, and solving
for perturbation pressure give the diagnostic equation for nonhydrostatic pressure
as
2
p
a
g
c
v.d
c
p.d

a
p
a
p
a
[
a
(v
) v]
[
a
f k v]
2
z
p
a
+ g

z

a
K
m
) v
(5.8)
(e.g., Pielke 1984). Since this equation is diagnostic, it does not have a time deriva-
tive. The pressure is calculated from other variables in a manner that ensures that
the continuity equation given in (5.7) is satised.
When pressure is calculated from(5.8), vertical scalar velocities can be extracted
diagnostically from (5.7), since u and v are known from the horizontal momentum
equations and
a
is available from (5.2). In sum, in an anelastic nonhydrostatic
model, the vertical momentumequation is not solved directly. Instead, a diagnostic
equation for nonhydrostatic pressure, whose solution is computationally intensive,
is solved.
5.1.4 Boussinesq approximation
The Boussinesq approximation is a special case of the anelastic approximation,
valid for cases where the vertical scale of circulation is much less than the scale
height of the atmosphere (shallow circulation). With the Boussinesq approxi-
mation, density (
a
) is replaced with a constant boundary-layer-averaged value
(
a
) in all the equations of atmospheric dynamics, except in the buoyancy term
of the vertical momentum equation. This differs from the anelastic approxima-
tion, in which density is horizontally averaged but varies for each layer in the
vertical. Because density is constant everywhere with the Boussinesq approxi-
mation, the continuity equation for air is nondivergent and given by
v = 0
rather than by (5.7). With the Boussinesq approximation, the vertical momen-
tum equation is the same as in (5.4), but with the eddy diffusion term added
back in, and the diagnostic equation for nonhydrostatic pressure is the same as
in (5.8), except that the second term on the left side of the equation is elim-
inated. The resulting equation is a Poisson partial differential equation. Lilly
(1996) discusses advantages and disadvantages of the anelastic and Boussinesq
approximations.
142
5.3 Pressure coordinate
z
5
z
5
z
6
z
4
z
3
z
2
z
1
Figure 5.1 Heights of altitude coordinate surfaces.
In the following sections, the conversion of the equations of atmospheric dynam-
ics from the altitude coordinate to the pressure, sigma-pressure, and sigma-altitude
coordinates is discussed. Whereas the altitude and sigma-altitude coordinates are
useful for solving hydrostatic and nonhydrostatic model equations, the pressure
and sigma-pressure coordinates are used for solving hydrostatic model equations
only. An additional vertical coordinate, the isentropic coordinate, in which layer
tops and bottoms are dened as surfaces of constant potential temperature, is dis-
cussed in Kasahara (1974).
5.2 ALTITUDE COORDINATE
Equations describing basic atmospheric processes were written in Chapters 3 and
4 in Cartesian and spherical horizontal coordinates and the altitude vertical coor-
dinate. In the altitude (z) coordinate, layer tops and bottoms are dened as surfaces
of constant altitude, and pressure varies in the x- and y-directions along these sur-
faces. Altitude is an independent variable, and pressure is a dependent variable.
This coordinate causes boundary-condition problems when surface elevations are
not uniform. In such cases, planes of constant altitude intercept topography, as
shown in Fig. 5.1. Special boundary conditions are needed in this situation.
The altitude coordinate can be used to simulate hydrostatic or nonhydro-
static model equations. In both cases, some variables are solved for prognostically
whereas others are solved for diagnostically. A prognostic equation is one in which
a time derivative is solved. The thermodynamic energy equation, species continuity
equation, and horizontal momentum equations are prognostic equations for the
potential virtual temperature, species mixing ratio, and horizontal scalar velocities,
respectively. A diagnostic equation is one in which a time derivative is not solved.
The equation of state is a diagnostic equation for any parameter within it for
hydrostatic and nonhydrostatic models. The hydrostatic equation is a diagnostic
equation in a hydrostatic model.
5.3 PRESSURE COORDINATE
In the pressure (p or isobaric) coordinate, layer tops and bottoms are dened
as surfaces of constant pressure. Since altitude is a function of pressure in the
143
p
a
,2
p
a
,1
p
a
,3
p
a
,4
p
a
,5
p
a,6
p
a
,1
p
a
,1
p
a,top
Figure 5.2 Heights of pressure coordinate surfaces. Each line is
a surface of constant pressure. Pressure thicknesses of each layer
are constant throughout the layer.
x
z
p
1
p
2
z
1
z
2
x
1
x
2
q
3
q
2
q
1
Figure 5.3 Intersection of pressure (p) with altitude (z) sur-
faces. The moist-air mass mixing ratio (q) varies along each
surface of constant pressure.
x- and y-directions, pressure is the independent variable and altitude is a depen-
dent variable. The pressure coordinate does not erase problems associated with
surface topography since surfaces of constant pressure intercept topography in the
pressure coordinate (Fig. 5.2). Equations in the pressure coordinate are derived
from the hydrostatic assumption. This assumption removes vertically propagat-
ing acoustic waves as a solution to the equations of atmospheric dynamics but
reduces the accuracy of the solution when horizontal grid cells are less than 3 km
in width. An advantage of the pressure coordinate is that the continuity equa-
tion for air in the coordinate requires neither a density nor a time-derivative
term. In this section, hydrostatic model equations for the pressure coordinate are
derived.
5.3.1 Gradient conversion from altitude to pressure coordinate
The gradient conversion from the altitude to the pressure coordinate can be
obtained diagrammatically. Figure 5.3 shows two lines of constant pressure
(isobars) intersecting two lines of constant altitude in the xz plane, where x is
a Cartesian coordinate distance. Suppose the moist-air mass mixing ratio q varies
144
along each surface of constant pressure, as shown in the gure. In this case, q varies
with distance as
q
2
q
3
x
2
x
1
=
q
1
q
3
x
2
x
1
+
p
2
p
1
x
2
x
1
q
1
q
2
p
1
p
2
(5.9)
which is an exact equivalence.
As x
2
x
1
0 and p
1
p
2
0, the differences in (5.9) can be approximated
by differentials. From Fig. 5.3, the approximations for each term are
q
x
z
=
q
2
q
3
x
2
x
1
q
x
p
=
q
1
q
3
x
2
x
1
p
a
x
z
=
p
2
p
1
x
2
x
1
q
p
a
x
=
q
1
q
2
p
1
p
2
(5.10)
Substituting these terms into (5.9) gives
q
x
z
=
q
x
p
+
p
a
x
q
p
a
x
(5.11)
which generalizes for any variable as
z
=
p
+
p
a
x

p
a
x
(5.12)
A similar equation can be written for the y-direction. Combining the x- and
y-equations in vector form yields the horizontal gradient conversion from
Cartesian-altitude to Cartesian-pressure coordinates as
z
=
p
+
z
( p
a
)

p
a
(5.13)
where
z
is the horizontal gradient operator in Cartesian-altitude coordinates given
in (4.81), and
p
= i
p
+j
p
(5.14)
is the horizontal gradient operator in Cartesian-pressure coordinates. The gradient
operator in Cartesian-pressure coordinates is used to take partial derivatives of
a variable along a surface of constant pressure. The derivative with respect to
pressure in (5.13) is taken in a xed column in the xy plane.
In (5.12), any dimension can be substituted for x. Substituting time t for x gives
the time-derivative conversion between the altitude and the pressure coordinate
as
z
=
p
+
p
a
t

p
a
t
(5.15)
145
Equation (5.13) may be used to obtain the relationship between the variation
of geopotential along a surface of constant pressure and the variation of pressure
along a surface of constant altitude. From (4.48), the geopotential was dened as
+ = gz. The geopotential is constant along surfaces of constant altitude (
z
+ = 0).
Substituting + = gz, p
a
z =
a
g, and +z = g into (5.13) and rearranging
give
z
( p
a
) =
p
a
+
p
+ =
p
a
gz
p
+ =
a
p
+ (5.16)
which relates changes in geopotential along surfaces of constant pressure to changes
in pressure along surfaces of constant altitude. This equation expands to
p
a
x
z
=
a
+
x
p
a
y
z
=
a
+
y
p
(5.17)
5.3.2 Continuity equation for air in the pressure coordinate
Equation (3.20) gave the continuity equation for air in Cartesian-altitude coordi-
nates. Dividing (3.20) into horizontal and vertical components gives
a
t
z
=
a
z
v
h
+
w
z
(v
h
z
)
a
w
a
z
(5.18)
where the subscript z indicates that the value is taken along a surface of constant
altitude, and v
h
= iu +jv. Applying (5.13) to v
h
gives
z
v
h
=
p
v
h
+
z
( p
a
)
v
h
p
a
(5.19)
Substituting (5.19) and dz = dp
a
a
g into (5.18) gives
a
t
z
=
a
p
v
h
+
z
( p
a
)
v
h
p
a
(v
h
z
)
a
+
a
g
(w
a
)
p
a
(5.20)
The vertical scalar velocity in the pressure coordinate (hPa s
1
) is dened as
w
p
=
dp
a
dt
=
p
a
t
z
+(v
) p
a
=
p
a
t
z
+(v
h
z
) p
a
+w
p
a
z
(5.21)
Substituting p
a
z =
a
g into (5.21) yields
w
p
=
a
g
z
t
z
+(v
h
z
) p
a
w
a
g (5.22)
If horizontal and temporal variations in pressure are ignored, the vertical scalar
velocity in the pressure coordinate simplies to w
p
= w
a
g. Thus, a positive ver-
tical scalar velocity in the pressure coordinate corresponds to downward motion,
or a negative vertical scalar velocity in the altitude coordinate. Horizontal and
temporal variations in pressure cannot be excluded when the continuity equation
146
for air is derived. Taking the partial derivative of w
p
from (5.22) with respect to
altitude gives
w
p
z
= g
a
t
z
+
z
( p
a
)
v
h
z
+(v
h
z
)
p
a
z
g
(w
a
)
z
(5.23)
Substituting the hydrostatic equation throughout yields
a
w
p
p
a
=
a
t
z
+
a
z
( p
a
)
v
h
p
a
+(v
h
z
)
a
a
g
(w
a
)
p
a
(5.24)
Adding (5.20) to (5.24) and compressing give the continuity equation for air in
Cartesian-pressure coordinates as
p
v
h
+
w
p
p
a
= 0 (5.25)
which expands to
u
x
+
v
y
p
+
w
p
p
a
= 0 (5.26)
Equation (5.25) shows that, in the pressure coordinate, the continuity equation for
air does not depend on air density, nor does it require a time-derivative term.
Example 5.1
Assume a grid cell in the pressure coordinate has dimension Lx = 5 km,
Ly = 5 km, and Lp
a
= 10 hPa, and that west, east, south, north, and
lower boundary scalar velocities are u
1
= 3, u
2
= 1 m s
1
, v
3
= +2, v
4
=
2 m s
1
, and w
p.5
= +0.02 hPa s
1
, respectively. Use the continuity equation
in the pressure coordinate to estimate the pressure-coordinate vertical scalar
velocity at the cells top.
SOLUTION
Applying the velocities and incremental distances to (5.26) gives
(1 +3) ms
1
5000 m
+
(2 2) ms
1
5000 m
+
(w
p.6
0.02) hPa s
1
10 hPa
= 0
which has solution w
p.6
= +0.016 hPa s
1
.
5.3.3 Total derivative in the pressure coordinate
The total derivative in the pressure coordinate may be used to derive the species
continuity equation, the thermodynamic energy equation, and the momentum
equations in the pressure coordinate. The total derivative in Cartesian-altitude
147
coordinates was
d
dt
=
z
+(v
h
z
) +w

z
(5.27)
Substituting time and horizontal gradient conversions from (5.15) and (5.13),
respectively, into (5.27) gives
d
dt
=
p
+
p
a
t
p
a
+(v
h
p
) +[(v
h
z
) p
a
]

p
a
+w

z
(5.28)
From (5.21), the vertical scalar velocity in the altitude coordinate can be written
as a function of the vertical scalar velocity in the pressure coordinate as
w =
p
a
t
z
+(v
h
z
) p
a
w
p
a
g
(5.29)
Substituting this equation and the hydrostatic equation into (5.28) and simplifying
give the total derivative in Cartesian-pressure coordinates as
d
dt
=
p
+(v
h
p
) +w
p
p
a
(5.30)
5.3.4 Species continuity equation in the pressure coordinate
Applying the total derivative in Cartesian-pressure coordinates to (3.54) gives the
species continuity equation in Cartesian-pressure coordinates as
dq
dt
=
q
t
p
+(v
h
p
)q +w
p
q
p
a
=
(
a
K
h
) q
a
+
N
e.t
n=1
R
n
(5.31)
If the concentration of a species is given as molecules per cubic centimeter of air (N),
particles per cubic centimeter of air (n
i
), or cubic centimeters of particle component
per cubic centimeter of air (v
q.i
), it can be converted to the moist-air mass mixing
ratio for use in (5.31) with
q =
Nm
a
A
q =
n
i
a
q =
v
q.i
a
(5.32)
respectively, where m is molecular weight (g mol
1
), A is Avogadros number
(molec. mol
1
),
i
is the volume of one particle in size bin i (cm
3
), and
p
is particle
mass density (g cm
3
).
Since R
n
does not include spatial derivatives, it does not need to be transformed
to the pressure coordinate. The eddy diffusion termin (5.31) includes spatial deriva-
tives. Its transformation is performed by breaking the terminto directional compo-
nents, applying the gradient conversion from (5.13), and applying dz = dp
a
a
g.
148
The result is
(
a
K
h
) q =
a
K
h.xx
q
x
a
K
h.yy
q
y
+

z
a
K
h.zz
q
z
p
+
p
a
x
p
a
a
K
h.xx
q
x
p
+
p
a
x
z
q
p
a
p
+
p
a
y
p
a
a
K
h.yy
q
y
p
+
p
a
y
z
q
p
a
+
a
g
2

p
a
2
a
K
h.zz
q
p
a
(5.33)
5.3.5 Thermodynamic energy equation in the pressure coordinate
Applying the total derivative in Cartesian-pressure coordinates to (3.76) gives the
thermodynamic energy equation in Cartesian-pressure coordinates as
v
t
p
+(v
h
p
)
v
+w
p
v
p
a
=
(
a
K
h
)
v
a
+

v
c
p.d
T
v
N
e.h
n=1
dQ
n
dt
(5.34)
The eddy diffusion terms here are the same as those used in the species continuity
equation, except that
v
is used here instead of q.
5.3.6 Horizontal momentum equations in the pressure coordinate
Applying the total derivative in Cartesian-pressure coordinates to velocity, neglect-
ing viscous terms, and substituting
z
( p
a
) =
a
p
+ from (5.16) into (4.71) give
the horizontal momentum equation in Cartesian-pressure coordinates as
v
h
t
p
+(v
h
p
)v
h
+w
p
v
h
p
a
= f k v
h
p
++
(
a
K
m
) v
h
a
(5.35)
The eddy diffusion term expands to (
a
K
m
)v
h
= i (
a
K
m
) u +
j (
a
K
m
) v. Applying the gradient conversion from (5.13) to the u-term gives
(
a
K
m
) u =
p
+
p
a
x
p
a
a
K
m.xx
u
x
p
+
p
a
x
z
u
p
a
p
+
p
a
y
p
a
a
K
m.yx
u
y
p
+
p
a
y
z
u
p
a
+
a
g
2

p
a
2
a
K
m.zx
u
p
a
(5.36)
149
The equation for the v-direction is similar. The three terms on the right side of
(5.36) represent accelerations in the x-direction due to westeast, southnorth,
and vertical turbulent transport, respectively, of westeast momentum.
5.3.7 Vertical momentum equation in the pressure coordinate
The vertical momentumequation in the pressure coordinate is the hydrostatic equa-
tion, since the conversion from the altitude to the pressure coordinate requires
the hydrostatic assumption. Over large horizontal scales (>3 km), the hydrostatic
assumption is reasonable. Over smaller scales and in the presence of large-scale con-
vective clouds, vertical accelerations are important, and the hydrostatic assumption
is not so good. In such cases, the altitude or sigma-altitude coordinate is generally
used, and a vertical momentum equation that includes a local acceleration term is
solved.
When the hydrostatic assumption is used, the vertical momentum equation in
the altitude coordinate is p
a
z =
a
g. Substituting g = +z from (4.48),
p
a
=
a
R
T
v
from (2.36), and T
v
=
v
P from (2.99) into the hydrostatic equation
gives the hydrostatic equation in the pressure coordinate as
+
p
a
=
R
T
v
p
a
=
R
v
P
p
a
=
R
v
p
a
p
a
1000 hPa
(5.37)
Substituting = R
c
p.d
into (5.37) and rearranging give another formof the equa-
tion as
d+ = c
p.d
v
d
p
a
1000 hPa
= c
p.d
v
dP (5.38)
A nite-difference discretization of this equation is given in Chapter 7.
5.3.8 Geostrophic wind in the pressure coordinate
Substituting (5.17) into (4.79) gives geostrophic scalar velocities in Cartesian-
pressure coordinates as
v
g
=
1
f
+
x
p
u
g
=
1
f
+
y
p
(5.39)
The vector form of (5.39) is
v
g
= iu
g
+jv
g
= i
1
f
+
y
p
+j
1
f
+
x
p
=
1
f
(k
p
+) (5.40)
which indicates that the geostrophic wind ows parallel to lines of constant geopo-
tential. Lines of constant geopotential (contour lines) on a constant-pressure surface
are analogous to lines of constant pressure (isobars) on a constant-altitude surface.
150
5.4 Sigma-pressure coordinate
500 hPa
surface
Contour line
Contour line
South
North
West
East
510 hPa
5.5 km
500 hPa
5.5 km
500 hPa
5.6 km
Figure 5.4 The geostrophic wind (arrow) ows parallel to contour lines on
a constant-pressure surface. In this gure, contour lines at 5.5 and 5.6 km
are shown on a 500-hPa surface. The resulting geostrophic wind originates
from the west.
Since the geostrophic wind ows parallel to isobars, it also ows parallel to contour
lines. Figure 5.4 shows two westeast contour lines at 5.5 and 5.6 km on a surface
of constant pressure. The geostrophic wind in this case ows from the west.
5.4 SIGMA-PRESSURE COORDINATE
A vertical coordinate in which surfaces of the coordinate variable do not intersect
ground topography is the sigma-pressure (-p) coordinate (Phillips 1957). In this
coordinate, layer tops and bottoms are dened as surfaces of constant , where
equals the difference between the layer bottom and model top pressures divided
by the difference between the model bottom and model top pressures. Since the
hydrostatic assumption is used to derive equations in the sigma-pressure coordi-
nate, vertically propagating acoustic waves are ltered out as a solution to these
equations. Nonhydrostatic ows are generally not simulated in the sigma-pressure
coordinate. In this section, hydrostatic model equations for the sigma-pressure
coordinate are derived.
5.4.1 Denitions
In the sigma-pressure coordinate system, layer tops and bottoms are dened as
surfaces of constant , given by
=
p
a
p
a.top
p
a.surf
p
a.top
=
p
a
p
a.top
a
(5.41)
where p
a
is the air pressure at the altitude of interest, p
a.surf
is the model surface
pressure, p
a.top
is the model top pressure, and
a
= p
a.surf
p
a.top
is the pressure
difference between the model surface and top (-value). From (5.41), the pressure
at a -level is
p
a
= p
a.top
+
a
(5.42)
151
s
1
= 0
s
6
= 1
s
5
s
4
s
3
s
2
p
a,top
s
1,
p
a,1x
s
1,
p
a,1y
Figure 5.5 Heights of sigma-pressure coordinate surfaces. Each
layer has the same -thickness but a different pressure thickness.
Subscripts x and y denote two horizontal locations. Surface pres-
sures vary horizontally.
x
s
1
s
2
z
1
z
2
x
1
x
2
p
3
p
2
p
1
z
Figure 5.6 Intersection of -p with z surfaces and values
of pressure at each intersection.
The parameter is a fraction 1. At the model top, = 0. At the model surface,
= 1. Layer boundaries correspond to -values that are constant in time. The
model top pressure, p
a.top
, is constant in space and time along a boundary corre-
sponding to = 0. The model surface pressure, p
a.surf
, varies in space and time
along the surface corresponding to = 1. Since p
a.surf
varies, p
a
and
a
must also
vary in space and time. Figure 5.5 shows heights of constant- surfaces in the
sigma-pressure coordinate.
5.4.2 Gradient conversion from the altitude
to the sigma-pressure coordinate
The conversion between pressure in the altitude coordinate and pressure in the
sigma-pressure coordinate can be derived from Fig. 5.6. The gure shows the inter-
section of with z-surfaces and the values of pressure at each intersection point
on an xz plane. From the gure, the change in pressure per unit distance is
p
2
p
3
x
2
x
1
=
p
1
p
3
x
2
x
1
+
1
x
2
x
1
p
1
p
2
(5.43)
152
which is an exact equivalence. Substituting
p
a
x
z
=
p
2
p
3
x
2
x
1
p
a
x
=
p
1
p
3
x
2
x
1
(5.44)
z
=

2
1
x
2
x
1
p
a
x
=
p
1
p
2
2
into (5.43) gives
p
a
x
z
=
p
a
x
p
a
x
(5.45)
Asimilar expression exists for the y-direction. The vector sumof the two equations
is
z
( p
a
) =
( p
a
) +(
z
)
p
a
(5.46)
where
i

x
j

y
(5.47)
is the horizontal gradient operator in Cartesian-sigma-pressure coordinates.
Equation (5.46) can be generalized for any variable with
z
=
+
z
()

(5.48)
which is the gradient conversion from Cartesian-altitude to Cartesian-sigma-
pressure coordinates.
The gradient of dened in (5.41), along a surface of constant altitude is
z
= ( p
a
p
a.top
)
z
+

z
( p
a
)
a
=
z
(
a
) +

z
( p
a
)
a
(5.49)
Substituting (5.49) into (5.48) yields another form of the gradient conversion,
z
=
z
(
a
)

z
( p
a
)
(5.50)
5.4.3 Gradient conversion from the pressure to the sigma-pressure
coordinate
The gradient conversion from the pressure to the sigma-pressure coordinate is
derived from Fig. 5.7. The gure shows intersections of pressure, altitude, and
surfaces in an xz plane. Fromthe gure, the change in moist-air mass mixing ratio
over distance is
q
1
q
3
x
2
x
1
=
q
2
q
3
x
2
x
1
+
2
x
2
x
1
q
1
q
2
(5.51)
153
x
z
p
1
p
2
z
1
z
2
x
1
x
2
q
3
q
2
q
1
s
1
s
2
Figure 5.7 Intersection of p, z, and surfaces in an xz plane.
Moist-air mass mixing ratios (q) on constant-pressure surfaces
are also shown.
which is an exact equivalence. Making substitutions similar to those in (5.44) gives
q
x
p
=
q
x
x
(5.52)
A similar equation is written for the y-direction. Writing the x- and y-equations in
gradient operator form and generalizing for any variable give the gradient conver-
sion from Cartesian-pressure to Cartesian-sigma-pressure coordinates as
p
=
+
p
()

(5.53)
Taking the gradient of along a surface of constant pressure and noting that
p
( p
a
) = 0 and
p
( p
top
) = 0 yield
p
() = ( p
a
p
top
)
p
+

p
( p
a
p
top
)
a
=
p
(
a
) (5.54)
Since horizontal gradients of
a
are independent of the vertical coordinate sys-
tem,
p
(
a
) =
(
a
) =
z
(
a
). Substituting this expression and (5.54) into (5.53)
gives the gradient conversion from Cartesian-pressure to Cartesian-sigma-pressure
coordinates as
p
=
(
a
)

(5.55)
5.4.4 Continuity equation for air in the sigma-pressure coordinate
5.4.4.1 Derivation
The continuity equation for air in the sigma-pressure coordinate is derived from
the continuity equation for air in the pressure coordinate, the hydrostatic equation,
and the total derivative in the sigma-pressure coordinate. The partial derivative of
154
pressure with respect to is p
a
=
a
. Substituting this expression and the
gradient conversion from (5.55) into the continuity equation for air in the pressure
coordinate of (5.25) gives

v
h
(
a
)
v
h
+
1
a
w
p
= 0 (5.56)
An expression for w
p
is now needed.
The vertical scalar velocity in the sigma-pressure coordinate ( s
1
) is
=
d
dt
(5.57)
The relationship between

and w
p
is found by substituting p
a
= p
a.top
+
a
and
(5.57) into (5.21) to yield
w
p
=
dp
a
dt
=
d
a
dt
+
d
dt

a
=
d
a
dt
+

a
(5.58)
If
a
is assumed to be constant with time, and horizontal gradients of
a
are ignored,
(5.22) simplies to w
p
= w
a
g. Substituting this term into (5.58) with constant
a
yields

= w
a
g
a
. Thus, positive vertical velocities in the sigma-pressure
coordinate result in downward motions. Because
a
varies in time and space, (5.58)
is not simplied in the present derivation.
The total derivative in Cartesian-sigma-pressure coordinates is found by replac-
ing p
a
with and w
p
with

in (5.30). The result is
d
dt
=
+v
h
(5.59)
Substituting the total derivative of
a
into (5.58) gives
w
p
=
a
t
+(v
h
)
a
a
(5.60)
where
a
= 0. The partial derivative of this equation with respect to is
w
p
a
t
+(v
h
)
a
+
(
a
)
v
h
+
a
(5.61)
Substituting (5.61) into (5.56), canceling terms, and compressing the result yield
the continuity equation for air in Cartesian-sigma-pressure coordinates as
a
t
+

(v
h
a
) +
a
= 0 (5.62)
Substituting v
h
a
for v in the horizontal component of (4.11), substituting the result
into (5.62), and multiplying through by R
2
e
cos give the continuity equation for
155
air in spherical-sigma-pressure coordinates as
R
2
e
cos
a
t
e
(u
a
R
e
) +

(v
a
R
e
cos )
+
a
R
2
e
cos
= 0
(5.63)
5.4.4.2 Pressure and vertical scalar velocity from the
continuity equation for air
The time rate of change of air pressure and vertical scalar velocity can be calculated
from the continuity equation for air. The integral of (5.62) over all -layers is
1
0
a
t
d =

1
0
(v
h
a
)d
a
0
0
d
(5.64)
The integral limits are = 0 and

= 0 at the model top and = 1 and

= 0 at
the model bottom. At the ground, vertical scalar velocities are always zero, and
at the model top, velocities in the -p coordinate are assumed to be zero to close a
boundary condition. In reality,

= 0 at the model top. Integrating (5.64) gives a
prognostic equation for column pressure in Cartesian-sigma-pressure coordinates
as
a
t
1
0
(v
h
a
)d (5.65)
The analogous equation in spherical-sigma-pressure coordinates is
R
2
e
cos
a
t
1
0
e
(u
a
R
e
) +

(v
a
R
e
cos )
d (5.66)
A numerical solution to this equation is shown in Chapter 7.
Once (5.66) has been written, an equation for the vertical scalar velocity at any
altitude can be derived. Integrating (5.62) with respect to from the model top to
any -level, and rearranging give
0
d

0
(v
h
a
)d
a
t
d (5.67)
Integrating (5.67) yields
a
=

0
(v
h
a
)d
a
t
(5.68)
which is the vertical scalar velocity at any -level. This equation is solved after
(
a
t)
has been obtained from (5.65). Applying the spherical-coordinate trans-

formation from (4.11) to (5.68) and multiplying through by R
2
e
cos give a
156
diagnostic equation for vertical scalar velocity at any -level as
a
R
2
e
cos =
e
(u
a
R
e
) +

(v
a
R
e
cos )
d R
2
e
cos
a
t
(5.69)
5.4.5 Species continuity equation in the sigma-pressure coordinate
Applying the total derivative from (5.59) to (3.54) gives the species continuity
equation in Cartesian-sigma-pressure coordinates as
dq
dt
=
q
t
+(v
h
)q +

=
(
a
K
h
)q
a
+
N
e.t
n=1
R
n
(5.70)
The eddy diffusion term expands in Cartesian-sigma-pressure coordinates to
(
a
K
h
)q =
a
K
h.xx
q
x
z
q
a
K
h.yy
q
y
z
q
a
g
2
2
a
2
a
K
h.zz
q
(5.71)
The vertical term was found by substituting p
a
=
a
into the vertical term of
(5.33).
Multiplying (5.70) by
a
, multiplying the continuity equation for air from(5.62)
by q, and summing the two equations give the ux-formspecies continuity equation
as
(
a
q)
t
+

(v
h
a
q) +
a
(
q)
=
a
a
K
h
) q
a
+
N
e.t
n=1
R
n
(5.72)
Applying the spherical coordinate transformation from (4.11) to (5.72) and mul-
tiplying through by R
2
e
cos give the ux form of the species continuity equation
in spherical-sigma-pressure coordinates as
R
2
e
cos

t
(
a
q)
e
(u
a
qR
e
) +

(v
a
qR
e
cos )
+
a
R
2
e
cos
q) =
a
R
2
e
cos
a
K
h
)q
a
+
N
e.t
n=1
R
n
(5.73)
157
5.4.6 Thermodynamic equation in the sigma-pressure coordinate
Applying the total derivative from(5.59) to (4.76) gives the thermodynamic energy
v
t
+(v
h
)
v
+

=
(
a
K
h
)
v
a
+

v
c
p.d
T
v
N
e.h
n=1
dQ
n
dt
(5.74)
Multiplying (5.74) by
a
, multiplying (5.62) by
v
, adding the two equations,
and compressing terms give the ux form of the thermodynamic energy equation
as
(
a
v
)
t
+

(v
h
v
) +
a
(
v
)
=
a
a
K
h
)
v
a
+

v
c
p.d
T
v
N
e.h
n=1
dQ
n
dt
(5.75)
The eddy diffusion term in this equation is the same as that in (5.71) except
v
is used in the present case instead of q. Applying the spherical coordinate
transformation from (4.11) to (5.75) and multiplying through by R
2
e
cos give
the ux form of the thermodynamic energy equation in spherical-sigma-pressure
coordinates as
R
2
e
cos

t
(
a
v
)
e
(u
a
v
R
e
) +

(v
a
v
R
e
cos )
+
a
R
2
e
cos
v
) =
a
R
2
e
cos
a
K
h
)
v
a
+

v
c
p.d
T
v
N
e.h
n=1
dQ
n
dt
(5.76)
5.4.7 Horizontal momentum equations
in the sigma-pressure coordinate
Applying the total derivative from (5.59) to the horizontal momentum equation in
Cartesian-altitude coordinates from (4.71) gives
v
h
t
+(v
h
)v
h
+

v
h
+ f k v
h
=
1
z
( p
a
) +
(
a
K
m
)v
h
a
(5.77)
The pressure-gradient term in (5.77) can be converted to the sigma-pressure coor-
dinate by combining (5.16) with (5.55). The result is
1
z
( p
a
) =
p
+ =
(
a
)
+
(5.78)
158
Substituting (5.78) into (5.77) gives the horizontal momentum equation in
Cartesian-sigma-pressure coordinates as
v
h
t
+(v
h
)v
h
+

v
h
+ f k v
h
=
++

(
a
)
+
+
(
a
K
m
)v
h
a
(5.79)
5.4.8 Coupling horizontal and vertical momentum equation
The hydrostatic equation is used in lieu of a prognostic vertical momentum equa-
tion in the sigma-pressure coordinate because the conversion of altitude to sigma-
pressure coordinate requires the hydrostatic assumption. Substituting p
a
=
a
R
T
v
and p
a
=
a
into (5.37) gives the hydrostatic equation in the sigma-pressure
coordinate as
+
a
R
T
v
p
a
=
a
=
a
a
(5.80)
Substituting (5.80) into (5.79) gives the horizontal momentum equation as
v
h
t
+(v
h
) v
h
+

v
h
= f k v
h
+
a
(
a
) +
(
a
K
m
) v
h
a
(5.81)
The term
a
should be modied to make the nite differencing of (5.81) in
Chapter 7 consistent with that of geopotential. Combining
a
= R
T
v
p
a
with R
=
c
p.d
, T
v
=
v
P, p
a
=
a
, and Pp
a
= Pp
a
gives
a
=
R
T
v
p
a
=
c
p.d
v
P
p
a
= c
p.d
v
P
p
a
=
c
p.d
a
P
(5.82)
Substituting (5.82) into (5.81), multiplying the result by
a
, multiplying (5.62) by
v
h
, and summing the two equations give the ux formof the horizontal momentum
(v
h
a
)
t
+v
h

(v
h
a
) +
a
(v
h
) v
h
+
a
v
h
)
=
a
f k v
h
+c
p.d
v
P
(
a
) +
a
(
a
K
m
) v
h
a
(5.83)
The advection terms in this equation expand to
v
h

(v
h
a
) = iu
(u
a
)
x
+
(v
a
)
y
+jv
(u
a
)
x
+
(v
a
)
y
(5.84)
a
(v
h
)v
h
= i
a
u
u
x
+v
u
y
+j
a
u
v
x
+v
v
y
(5.85)
159
Substituting these terms into (5.83), applying the transformation from (4.31),
applying (4.1), and multiplying through by R
2
e
cos give the ux forms of the
horizontal momentum equations in spherical-sigma-pressure coordinates as
R
2
e
cos

t
(
a
u)
e
(
a
u
2
R
e
) +

(
a
uvR
e
cos )
+
a
R
2
e
cos
u)
=
a
uvR
e
sin +
a
f vR
2
e
cos R
e
a
+
e
+c
p.d
v
P
+ R
2
e
cos
a
(
a
K
m
) u (5.86)
R
2
e
cos

t
(
a
v)
e
(
a
uvR
e
) +

(v
2
a
R
e
cos )
+
a
R
2
e
cos
v)
=
a
u
2
R
e
sin
a
f uR
2
e
cos R
e
cos
a
+
+c
p.d
v
P
+R
2
e
cos
a
(
a
K
m
) v (5.87)
The westeast eddy diffusion term expands in Cartesian-sigma-pressure coordi-
nates to
(
a
K
m
) u =
a
K
m.xx
u
x
z
u
a
K
m.yx
u
y
z
u
+

a
g
2
2
a
2
a
K
m.zx
u
(5.88)
The expression for (
a
K
m
) v is similar to (5.88), except v, K
m.xy
, K
m.yy
, and
K
m.zy
are used in the new term instead of u, K
m.xx
, K
m.yx
, and K
m.zx
, respectively.
5.5 SIGMA-ALTITUDE COORDINATE
The sigma-altitude (s-z) coordinate is dened such that layer tops and bottoms are
surfaces of constant s, where s equals the difference between the model top altitude
and the bottom altitude of a layer divided by the difference between the model
top and the model bottom altitude (e.g., Kasahara 1974). This coordinate is used
to simulate nonhydrostatic or hydrostatic ows. In the sigma-altitude coordinate,
layer thicknesses do not change, whereas in the sigma-pressure coordinate, layer
thicknesses change continuously. An advantage of the sigma-altitude and sigma-
pressure coordinates in comparison with the altitude and pressure coordinates is
that the former do not permit model layers to intercept ground topography, whereas
the latter do. In this section, elastic nonhydrostatic model equations for the sigma-
altitude coordinate are derived. Conversion to anelastic and hydrostatic equations
is also discussed.
160
5.5 Sigma-altitude coordinate
5.5.1 Denitions
Layer tops and bottoms in the sigma-altitude coordinate are dened as surfaces of
constant s, where s is dened as
s =
z
top
z
z
top
z
surf
=
z
top
z
Z
t
(5.89)
In this equation, z is the altitude at a layer bottom boundary, z
surf
is the model
bottom (surface) altitude, z
top
is the model top altitude, Z
t
= z
top
z
surf
is the
altitude difference between the model top and bottom, and s is a fraction 1. At
the model top, s = 0. At the model bottom, s = 1. From (5.89), the altitude at a
given s-level is
z = z
top
Z
t
s (5.90)
Figure 5.5, which shows heights of constant sigma-pressure surfaces, can also be
used to show heights of constant sigma-altitude surfaces by replacing s with
ss. In the s-z coordinate, layer tops and bottoms correspond to s-values that are
constant in time. Altitudes, including z, z
surf
, and z
top
, vary in space, but not in time,
along surfaces of constant s. Pressure (p
a
) varies in space and time along surfaces
of constant s.
5.5.2 Gradient conversion from altitude to sigma-altitude coordinate
The horizontal gradient conversion of a variable from Cartesian-altitude to
Cartesian-sigma-altitude coordinates is obtained in a manner similar to that for the
conversion fromCartesian-altitude to Cartesian-pressure coordinates. The result is
z
=
s
+
z
(s)

s
(5.91)
Taking the horizontal gradient of (5.89) and noting that
z
(z) =
z
(z
top
) = 0 give
z
(s) =
z
top
z
Z
2
t
z
(Z
t
) =
s
Z
t
z
(Z
t
) (5.92)
Substituting this result into (5.91) yields another form of the gradient conversion
as
z
=
s

s
Z
t
z
(Z
t
)

s
(5.93)
The time-derivative conversion from the altitude to sigma-altitude coordinate is
z
=
s
(5.94)
which was derived by substituting s for p in (5.15) and noting that (st)
z
= 0.
161
The vertical scalar velocity in the sigma-altitude coordinate (s s
1
) is
s =
ds
dt
= (v
h
z
) s +w
s
z
= (v
h
z
) s
w
Z
t
(5.95)
which has a zero local time derivative, since (st)
z
= 0. Equation (5.95) was
derived under the assumption that sz = 1Z
t
, obtained by differentiating
(5.89). A positive vertical scalar velocity in the sigma-altitude coordinate corre-
sponds to downward motion, just as in the sigma-pressure and pressure coor-
dinates. Substituting

s for w
p
and s for p in (5.30) gives the total derivative in
Cartesian-sigma-altitude coordinates as
d
dt
=
s
+(v
h
s
) +

s

s
(5.96)
5.5.3 Continuity equation for air in the sigma-altitude coordinate
The continuity equation for air in Cartesian-sigma-altitude coordinates is derived
from the continuity equation for air in Cartesian-altitude coordinates. Splitting
(3.20) into horizontal and vertical components and applying (5.91) to v
h
and
a
give
a
t
s
=
a
s
v
h
+
z
(s)
v
h
s
+
w
z
v
h
s
(
a
) +
z
(s)

a
s
a
z
(5.97)
Rewriting (5.95) as w = Z
t
[v
h
z
(s)

s] and differentiating give
w
z
= Z
t
z
(s)
v
h
z
+(v
h
z
)
s
z

s
z
(5.98)
Substituting dz = Z
t
ds into (5.98) yields
w
z
=

s
s

z
(s)
v
h
s
+
1
Z
t
(v
h
z
) Z
t
(5.99)
Substituting w = Z
t
[v
h
z
(s)

s], (5.99), and sz = 1Z
t
into (5.97)
gives
a
t
s
=
a
s
v
h
+

s
s
+
1
Z
t
(v
h
z
)Z
t
(v
h
s
)
a
a
s
(5.100)
162
Substituting
z
(Z
t
) =
s
(Z
t
) into (5.100) gives the elastic nonhydrostatic con-
tinuity equation for air in Cartesian-sigma-altitude coordinates as
a
t
s
=
1
Z
t
s
(v
h
a
Z
t
)

s
(
s
a
)
=
1
Z
t
(u
a
Z
t
)
x
+
(v
a
Z
t
)
y
s
a
)
s
(5.101)
When (5.101) is used to predict air density, vertically propagating acoustic waves
are one solution to the equations of atmospheric dynamics. To eliminate these
waves and enhance numerical stability, the anelastic approximation is made to
(5.101) by setting
a
t = 0 (Section 5.1).
If hydrostatic balance is assumed, vertically propagating acoustic waves are also
eliminated as a possible solution to the basic equations. Substituting sz = 1Z
t
into (2.40) gives the hydrostatic equation in the sigma-altitude coordinate as
a
=
1
g
p
a
z
=
1
Z
t
g
p
a
s
(5.102)
where primes have been added to indicate that density and pressure are in hydro-
static balance. Substituting (5.102) into (5.101) gives the hydrostatic continuity
equation in Cartesian-sigma-altitude coordinates as
a
s
=
s
v
h
p
a
s
s
p
a
s
(5.103)
5.5.4 Species continuity equation in the sigma-altitude coordinate
The species continuity equation in Cartesian-sigma-altitude coordinates is obtained
by applying the total derivative from (5.96) to (3.54). The result is
dq
dt
s
=
q
t
s
+(v
h
s
)q +

s
q
s
=
(
a
K
h
)q
a
+
N
e.t
n=1
R
n
(5.104)
The eddy diffusion term in (5.104) expands in Cartesian-sigma-altitude coordi-
nates to
(
a
K
h
)q =
s
+
s
x
a
K
h.xx
q
x
s
+
s
x
z
q
s
s
+
s
y
a
K
h.yy
q
y
s
+
s
y
z
q
s
+
1
Z
2
t
a
K
h.zz
q
s
(5.105)
where sz = 1Z
t
was used to obtain the vertical term.
163
5.5.5 Thermodynamic equation in the sigma-altitude coordinate
The thermodynamic energy equation in Cartesian-sigma-altitude coordinates is
obtained by applying (5.96) to (3.76). The result is
v
t
s
+(v
h
s
)
v
+

s
v
s
=
(
a
K
h
)
v
a
+

v
c
p.d
T
v
N
e.h
n=1
dQ
n
dt
(5.106)
Eddy diffusion is treated as in (5.105), except
v
is used in the new term instead
of q.
5.5.6 Horizontal momentum equations in the sigma-altitude
coordinate
Applying (5.96) to (4.71) gives the horizontal momentum equation in Cartesian-
sigma-altitude coordinates as
v
h
t
s
+(v
h
s
)v
h
+

s
v
h
s
+ f k v
h
=
1
z
( p
a
) +
(
z
a
K
m
z
)v
h
a
(5.107)
Applying the coordinate conversion from (5.93) to pressure gives
z
( p
a
) =
s
( p
a
)
s
Z
t
z
(Z
t
)
p
a
s
(5.108)
Substituting (5.108) into (5.107) gives the horizontal momentum equation as
v
h
t
s
+(v
h
s
)v
h
+

s
v
h
s
= f k v
h
s
( p
a
)
s
Z
t
z
(Z
t
)
p
a
s
(
a
K
m
)v
h
(5.109)
Equation (5.109) expands in the x- and y-directions to
u
t
+u
u
x
+v
u
y
s
+

s
u
s
= f v
1
p
a
x
s
Z
t
Z
t
x
z
p
a
s
(
a
K
m
)u
(5.110)
v
t
+u
v
x
+v
v
y
s
+

s
v
s
= f u
1
p
a
y
s
Z
t
Z
t
y
z
p
a
s
(
a
K
m
)v
(5.111)
164
respectively. The eddy diffusion term in the x-direction is
(
a
K
m
)u =
s
+
s
x
a
K
m.xx
u
x
s
+
s
x
z
u
s
s
+
s
y
a
K
m.yx
u
y
s
+
z
u
s
+
1
Z
2
t
a
K
m.zx
u
s
(5.112)
The expression for (
a
K
m
)v is similar.
5.5.7 Vertical momentum equation in the sigma-altitude coordinate
Substituting sz = 1Z
t
into the vertical momentum equation in Cartesian-
altitude coordinates from (4.75) and expanding the total derivative of w with
(5.96) give
w
t
+u
w
x
+v
w
y
s
+

s
w
s
= g +
1
Z
t
a
p
a
s
+
(
a
K
m
) w
a
(5.113)
Substituting w = Z
t
[v
h
z
(s)

s] from (5.95) into (5.113) gives the vertical
momentum equation in Cartesian-sigma-altitude coordinates as
s
+u
s
+v
s
+

s

s

Z
t
u
s
x
z
+ Z
t
v
s
y
z
Z
t
= g +
1
Z
t
a
p
a
s
+
1
a
(
a
K
m
)
Z
t
u
s
x
z
+ Z
t
v
s
y
z
Z
t
(5.114)
The diffusion term (
a
K
m
)(Z
t
s) is similar to (
a
K
m
)u from (5.112),
except that Z
t
s, K
m.xz
, K
m.yz
, and K
m.zz
are used in the new term instead of u,
K
m.xx
, K
m.yx
, and K
m.zx
.
5.5.8 Basic equations in spherical-sigma-altitude coordinates
The elastic nonhydrostatic sigma-altitude equations derived above can be converted
from Cartesian to spherical horizontal coordinates with (4.1), (4.11), and (4.31).
The resulting equations are given as follows:
Elastic continuity equation for air from (5.101)
R
2
e
cos
a
t
s
=
1
Z
t
e
(u
a
Z
t
R
e
) +

(v
a
Z
t
R
e
cos )
s
R
2
e
cos
s
(
s
a
)
(5.115)
Species continuity equation from (5.104)
q
t
+
u
R
e
cos
q
e
+
v
R
e
q
s
+

s
q
s
=
(
a
K
h
)q
a
+
N
e.t
n=1
R
n
(5.116)
165
Thermodynamic energy equation from (5.106)
v
t
+
u
R
e
cos
e
+
v
R
e
s
+

s
v
s
=
(
a
K
h
)
v
a
+

v
c
p.d
T
v
N
e.h
n=1
dQ
n
dt
(5.117)
Westeast momentum equations from (5.110)
u
t
+
u
R
e
cos
u
e
+
v
R
e
u
s
+

s
u
s
=
uv tan
R
e
+ f v
1
a
R
e
cos
p
a
s
Z
t
Z
t
z
p
a
s
+
(
a
K
m
) u
a
(5.118)
Southnorth momentum equation from (5.111)
v
t
+
u
R
e
cos
v
e
+
v
R
e
v
s
+

s
v
s
=
u
2
tan
R
e
f u
1
R
e
p
a
s
Z
t
Z
t
z
p
a
s
+
(
a
K
m
) v
a
(5.119)
Vertical momentum equation from (5.114)
s
+
u
R
e
cos
s
+
v
R
e
s
+

s

s
Z
t
u
R
e
cos
z
+
Z
t
v
R
e
z
Z
t
= g +
1
Z
t
a
p
a
s
+
1
a
(
a
K
m
)
Z
t
u
R
e
cos
z
+
Z
t
v
R
e
z
Z
t
(5.120)
As described in more detail in Section 5.1.2, these equations can be solved (a)
explicitly with a small time step in every term, (b) explicitly with a small time
step in terms producing acoustic waves and a large time step in all other terms, or
(c) implicitly or semiimplicitly with a large time step in terms producing acoustic
waves and explicitly with a large time step in all other terms.
Alternatively, the elastic equations above can be converted to anelastic form by
deriving a diagnostic equation for nonhydrostatic pressure (Section 5.1.3). In such a
case, the vertical momentum equation from (5.114) is not solved, since the vertical
scalar velocity is determined diagnostically from the anelastic continuity equation,
discussed following (5.101). In that case, the potential virtual temperature is still
found from the thermodynamic energy equation, the horizontal scalar velocities
are still found from the horizontal momentum equations, and the specic humidity
is still found from the species continuity equation.
166
5.7 Problems
5.6 SUMMARY
The equations of atmospheric dynamics include the continuity equation for air, the
species continuity equation, the thermodynamic energy equation, the horizontal
momentum equation, and the vertical momentum equation. In this chapter, the
vertical coordinate in the equations was transformed from the altitude to the pres-
sure, sigma-pressure, and sigma-altitude coordinates. With the altitude (and pres-
sure) coordinate, model layers may intersect surface topography. With the sigma-
altitude and sigma-pressure coordinates, model layers are terrain-following and
cannot intersect topography. When the pressure or sigma-pressure coordinate is
used, the atmosphere is assumed to be in hydrostatic balance. This assumption is
reasonable when coarse horizontal scales are simulated. For ner scales, the alti-
tude or sigma-altitude coordinate should be used so that nonhydrostatic motions
can be simulated.
5.7 PROBLEMS
5.1 (a) Assume Lx = 5 km, Ly = 5 km, and Lp
a
= 10 hPa for a grid cell
in the pressure coordinate and that the west, east, south, north, and
lower boundary scalar velocities are u
1
= 2 m s
1
, u
2
= +1 m s
1
, v
3
=
+1 m s
1
, v
4
=2 m s
1
, and w
5
=+0.03 m s
1
, respectively. Convert w
5
to the pressure coordinate with w
p
= w
a
g, assuming T = 284 K and
p
a
= 980 hPa. Use the pressure-coordinate continuity equation for air to
estimate w
p
at the top of the cell.
(b) For the same cell as in part (a), assume the west, east, south, north, lower,
and upper boundary mass mixing ratios of a gas are 0.004, 0.005, 0.003,
0.004, 0.0045, and 0.0055 kg kg
1
, respectively. Estimate the change in q
at the center of the cell after 500 s ignoring eddy diffusion and external
sources/sinks.
5.2 Assume a horizontal grid cell has dimension Lx = 500 km and Ly =
400 km, centered at = 30
N. Assume that the cell is on a surface of

constant pressure at p
a
= 500 hPa and that the altitudes of the west, east,
south, and north boundaries are 5.5, 5.4, 5.6, and 5.3 km, respectively.
Calculate geostrophic scalar velocities (u
g
and v
g
) and the geostrophic wind
speed.
5.3 Assume that a cell in the -p coordinate has dimension Lx = 4 km, Ly =
5 km, and L = 0.05. The west, east, south, and north boundary u- and
a
-values are u
1
= 2 m s
1
and
a,1
= 748 hPa, u
2
= +1 m s
1
and
a,2
=
752 hPa, v
3
= 1 m s
1
and
a,3
= 749 hPa, and v
4
= 2 m s
1
and
a,4
=
753 hPa, respectively. Assume p
a.top
= 250 hPa, = 0.9 at the cell bottom,
the grid-cell center
a
-value is an average of the four boundary values, T
v
=
298 K, the lower boundary vertical scalar velocity in the altitude coordinate
is w
5
= +0.02 m s
1
, and the air is dry.
(a) Convert vertical scalar velocity from the altitude to the sigma-pressure
coordinate with

= w
a
g
a
. Use the continuity equation for air in
the sigma-pressure coordinate to estimate the sigma-pressure coordinate
vertical scalar velocity at the top of the cell, assuming
a
t = 0.
167
(b) Assume that the west, east, south, north, lower, and upper boundary
values of
v
are 299, 297, 304, 301, 300, and 302 K, respectively, and that
no eddy diffusion or external sources/sinks exist. Estimate the value of
v
at the center of the grid cell after 200 s.
5.4 Assume that a horizontal grid cell has dimension Lx = 5 km and Ly
= 4 km and that
v
and p
a
on the west, east, south, and north boundaries of
the cell are
v.1
= 298 K and p
a,1
= 1010 hPa,
v.2
= 304 K and p
a,2
= 1004
hPa,
v.3
= 302 K and p
a,3
= 1000 hPa, and
v.4
= 301 K and p
a,4
= 1006 hPa,
respectively. What is the change in the u- and v-component scalar velocities
after 10 min due to the pressure gradient force alone? Assume the air is dry.
5.5 Derive the continuity equation for air in Cartesian-sigma-pressure coordi-
nates from the continuity equation for air in Cartesian-altitude coordinates.
5.6 Derive the horizontal momentumequationinCartesian-pressure coordinates
from that in Cartesian-sigma-pressure coordinates. Ignore conversion of the
eddy diffusion term.
5.7 Write a computer script to set up a model grid over the globe. Assume the
grid stretches from centered at 88
S to +88
N, where d = 4
and from
e
centered at 177.5
W to +177.5
E, where d
e
= 5
. Calculate dx and dy
at the southern and western boundaries, respectively, of each grid cell, and
print the values to a table.
168
6
Numerical solutions to partial
differential equations
A
tmospheri c models simulate physical processes described by ordinary and
partial differential equations. For example, gas and aqueous chemistry and gas-
to-particle conversion processes are described by ordinary differential equations,
and transport processes are described by partial differential equations. In this chap-
ter, ordinary and partial differential equations are dened, and numerical methods
of solving partial differential equations are discussed. Methods of solving partial
differential equations include nite-difference, series expansion, and nite-volume
methods. A special case of the nite-difference method is the semi-Lagrangian
method. Two series expansion methods are nite-element and pseudospectral meth-
ods. Below, several solution methods are applied to the advectiondiffusion equa-
tion, which is a unidirectional form of the species continuity equation. In addi-
tion, time-stepping schemes and their stability characteristics are discussed. Such
schemes include the Forward Euler, Implicit, CrankNicolson, Leapfrog, Matsuno,
Heun, AdamsBashforth, and RungeKutta schemes. Finally, necessary character-
istics of schemes that solve the advectiondiffusion equation in three-dimensional
models are discussed.
6.1 ORDINARY AND PARTIAL DIFFERENTIAL EQUATIONS
An ordinary differential equation (ODE) is an equation with one independent
variable, such as time, and a partial differential equation (PDE) is an equation
with more than one independent variable, such as time and space. ODEs and
PDEs are classied by their order and degree. The order is the highest derivative
rank of the equation, and the degree is the highest polynomial exponent of the
highest derivative. A homogeneous differential equation is an equation that does
not contain a terminvolving the independent variable. Alinear differential equation
is one in which the dependent variable and its derivatives do not appear in second-
degree or higher terms and in which the dependent variable is not multiplied by
other derivatives of itself.
Table 6.1 shows ordinary and partial differential equations of varying orders
and degrees. Table 6.1 equations (a), (b), (d), (e), and (f) are homogeneous, and the
remaining equations are inhomogeneous. Table 6.1 (b) and (e) are linear, and the
rest are nonlinear. Chemical equations are rst-order, rst-degree, homogeneous
ODEs (e.g., Table 6.1 (a) and (b)). These equations are either linear or nonlin-
ear. The species continuity equation and the thermodynamic energy equation are
169
Numerical solutions to partial differential equations
Table 6.1 Examples of the orders and degrees of ordinary and partial
Order, degree Ordinary differential equations Partial differential equations
First-order, rst-degree (a)
dN
dt
= 16 4N
2
(e)
N
t
+
(uN)
x
= 0
First-order, rst-degree (b)
dN
dt
= 3AB 4NC (f)
u
t
+u
u
x
+v
u
y
= 0
Second-order, rst-degree (c)
d
2
N
dt
2
+
dN
dt
+5t = 0 (g)

2
N
t
2
+

2
N
x
2
= 3t
2
+ x
Second-order, second-degree (d)
d
2
N
dt
2
2
+
dN
dt
+4 = 0 (h)
2
N
t
2
2
+
N
x
= t x
The variable t is time, x is westeast distance, y is southnorth distance, N, A, B, and C are
concentrations, u is westeast scalar velocity, and v is southnorth scalar velocity.
rst-order, rst-degree, homogeneous, linear PDEs (e.g., Table 6.1 (e)). The momen-
tum equation is a rst-order, rst-degree, homogeneous, nonlinear PDE (e.g.,
Table 6.1 (f)).
Boundary conditions for ODEs and PDEs must be specied. When conditions
are known at one end of a domain but not the other, an initial value problemarises.
If concentrations (N) are known at time t = 0, if time is the independent variable,
and if concentration is the dependent variable, then the solution to a set of ODEs
is an initial value problem. When conditions are known at both ends of a domain,
the solution to a set of ODEs is a boundary value problem. If time and westeast
direction (x) are independent variables, if concentration is the dependent variable,
and if the concentrations are known everywhere at t = 0 and at both ends of the
spatial domain at all times, the solution to a set of PDEs is an initial value problem
with respect to time and a boundary value problem with respect to space.
6.2 OPERATOR SPLITTING
Major processes in an atmospheric model are often solved separately from each
other. Suppose a model treats dynamics, transport, and gas chemistry. Each of
these processes may be solved sequentially during a common time interval with a
unique numerical scheme that takes a unique number of time steps. A time step
is an increment in time for a given process. A time interval is the period during
which several time steps of a process are solved without interference by another
process. Suppose the time step for dynamics is 6 s, that for transport is 300 s, that for
chemistry is variable, and the time interval common to all processes is 300 s. During
the time interval, 50 dynamics time steps are taken, followed by 1 transport time
step, followed by a variable number of chemistry time steps. After the dynamics time
interval, the resulting wind speeds are used as inputs into the transport calculation.
During the transport time interval, which equals the transport time step, gases are
170
6.3 Advectiondiffusion equations
Dynamics
Transport
Gas chemistry
Time interval 1
Dynamics
Transport
Gas chemistry
Time interval 2
Figure 6.1 Example of operator-splitting scheme. During the rst
time interval, dynamics, transport, and gas chemistry are solved
sequentially. Values determined from the end of a time interval
after one process are used to initialize values at the beginning of
the same time interval for another process. Values from the end of
the last process in one time interval are used at the beginning of the
rst process in the next time interval.
moved around the grid. Final concentrations from the transport time interval are
used as initial values for the rst chemistry time step of the gas chemistry time
interval. Final values from the chemistry time interval are used as initial values for
the rst dynamics time step in the next dynamics time interval. Figure 6.1 illustrates
this example.
The isolation of individual processes during a time interval is called time splitting
or operator splitting. Operator splitting is used because computers today cannot
solve all model ODEs and PDEs simultaneously in three dimensions. Yanenko
(1971) discusses the theoretical basis behind operator splitting with respect to
certain mathematical equations.
6.3 ADVECTIONDIFFUSION EQUATIONS
First-order, rst-degree, homogeneous, linear or nonlinear partial differential equa-
tions solved in atmospheric models include the species continuity equation, the
thermodynamic energy equation, and the directional momentum equations. In
Chapter 7, a method of solving these equations together is given. In this chapter,
methods of solving advectiondiffusion equations, which are operator-split forms
of the species continuity equation, are discussed. Advectiondiffusion equations are
derived by considering that the four-dimensional (t, x, y, z) species continuity equa-
tion can be divided into three two-dimensional partial differential equations ([t, x],
[t, y], and [t, z]) and a single one-dimensional (t) ordinary differential equation.
The sequential solution to the four operator-split equations approximates the solu-
tion to the original four-dimensional equation. This method of operator splitting
a mathematical equation is called the locally one-dimensional (LOD) procedure
or the method of fractional steps (e.g., Yanenko 1971; Mitchell 1969) and has
been used widely in atmospheric models (e.g., Reynolds et al. 1973; Carmichael
et al. 1986; Toon et al. 1988).
Fromthe four-dimensional species continuity equation given in (3.52), the west
east, southnorth, and vertical unidirectional advectiondiffusion equations can be
171
written in number concentration units as
N
t
+
(uN)
x

x
K
h.xx
N
x
= 0 (6.1)
N
t
+
(vN)
y

y
K
h.yy
N
y
= 0 (6.2)
N
t
+
(wN)
z

z
K
h.zz
N
z
= 0 (6.3)
respectively. The solution order of these equations may be reversed each time inter-
val to improve accuracy (e.g., Yanenko 1971). Thus, if the equations are solved in
the order (6.1), (6.2), (6.3) during one time interval, they may be solved in the order
(6.3), (6.2), (6.1) during the next time interval. Fractional-step schemes associated
with order reversal are called alternating-directions schemes.
The remaining terms in the species continuity equation are external source/sink
terms. These terms may be operator-split from the advectiondiffusion equations
as a single ordinary differential equation
N
t
=
N
e.t
n=1
R
n
(6.4)
or split into several ODEs. Equation (6.4) can be solved before or after (6.1)(6.3)
are solved.
In moist-air mass-mixing-ratio units, the operator-split westeast advection
diffusion equation can be written from (3.54) as
q
t
+u
q
x

1
a
K
h.xx
q
x
= 0 (6.5)
Analogous equations can be written for the southnorth and vertical directions
and for external sources/sinks. In the following subsections, nite-difference, series
expansion, and nite-volume methods of approximating derivatives and of solving
advectiondiffusion equations are discussed.
6.4 FINITE-DIFFERENCE APPROXIMATIONS
Approximate solutions to partial differential equations, such as advection
diffusion equations, can be found with nite-difference, series expansion, or nite-
volume methods. The purpose of using an approximation is to reduce the solution
space for each continuous differential function from an innite to a nite number
of spatial or temporal nodes in order to speed up computation of the differential
equation.
A nite-difference approximation involves the replacement of each continuous
differential operator (d) with a discrete difference analog (L). This analog is an
approximation written in terms of a nite number of values of the variable being
operated on at each temporal or spatial node. If the westeast scalar velocity is a
continuous function in space at a given time, its values can be mapped from the
172
6.4 Finite-difference approximations
u
x
u
x
i1 i i+1
u
i1
u
i
u
i+1
u
i+2
x
i1
x
i
x
i+1
x
i+2
Figure 6.2 Discretization of a continu-
ous westeast scalar velocity u
x
. The
westeast grid is broken into discrete
cells, and u-values are mapped from the
continuous function to the edge of each
cell. The arrows in the cells represent
magnitudes of the wind speed. Distances
along the x-axis are also mapped to the
cells.
function to a discretized westeast grid, as shown in Fig. 6.2. The grid consists of
several grid cells (also called grid boxes, grid points, or nodes) placed any distance
apart.
Table 6.1 (e) and (f) show partial differential equations commonly simulated in
atmospheric models. Such equations are written with respect to time and space. The
solution to Table 6.1 (e) requires nite-difference analogs for Nt and (uN)x.
The solution to Table 6.1 (f) requires nite-difference analogs for ut, ux,
and uy.
6.4.1 Consistency, convergence, and stability
A numerical solution can replicate an exact solution to a partial differential equa-
tion if several criteria are met. First, a nite-difference analog in space or time must
converge to its differential expression when terms in the analog approach zero.
For example, if LNLx is a nite-difference analog of Nx, the convergence
condition
N
x
= lim
Lx0
LN
Lx
(6.6)
must be met for the approximation to be accurate.
Second, a nite-difference analog must be consistent. The nite-difference ana-
log LNLx in (6.6) is obtained from a Taylor series expansion. In the expansion,
high-order terms are neglected to reduce the computational burden of the approx-
imation. The difference between the full Taylor series expansion and the trun-
cated approximation is the truncation error. A nite-difference approximation of
173
a derivative is consistent if the truncation error of the approximation approaches
zero as Lx (or Lt) approaches zero. Consistency occurs when
lim
Lx0
TE
LN
Lx
= 0 (6.7)
where TE is the truncation error of the approximation LNLx.
Third, if a nite-difference approximation is consistent, the rate at which its
truncation error approaches zero depends on the order of approximation. The
order of approximation is the order of the lowest-order term in the Taylor series
expansion neglected in the approximation. The higher the order of approximation,
the faster the truncation error converges toward zero upon an increase in spatial (or
temporal) resolution. Thus, with the same Lx, a high-order approximation is more
accurate than a low-order approximation. For the same truncation error, a low-
order approximation requires a smaller Lx than does a high-order approximation.
In sum, a high-order approximation with a large Lx can have the same trunca-
tion error as a low-order approximation with a small Lx. Because a high-order
approximation includes more terms, it requires more computations than does a
low-order approximation with the same Lx. Obtaining high order with respect to
one variable, such as space, is useful only if the order of the other variable, such as
time, is also high. Otherwise, low accuracy in the time derivative swamps the high
accuracy in the space derivative. An optimal nite-difference solution has similar
order in space and time.
Fourth, while individual nite-difference analogs must converge toward exact
differentials, the overall numerical solution to a PDE must converge to an exact
solution when spatial and temporal differences decrease toward zero. If N
e.x.t
is
an exact solution, and N
f.x.t
is a nite-difference approximation of a PDE, overall
convergence occurs when
lim
Lx.Lt0
N
e.x.t
N
f.x.t
= 0 (6.8)
If a numerical solution is nonconvergent, it is not useful.
Fifth, for a numerical method to be successful, it must be stable. Stability occurs
if the absolute-value difference between the numerical and exact solutions does not
grow over time. Thus,
lim
t
N
e.x.t
N
f.x.t
C (6.9)
where C is a constant. Stability often depends on the time-step size used. If a
numerical solution is stable for any time step smaller than a specied value, the
solution is conditionally stable. If a solution is stable, regardless of the time step,
it is unconditionally stable. If a solution is unstable, regardless of the time step, it
is unconditionally unstable.
A scheme that is unconditionally unstable cannot be convergent overall, but
individual nite-difference analogs in an unstable scheme may converge and may
be consistent. In other words, consistency and convergence of individual analogs
174
do not guarantee stability. On the other hand, stability is guaranteed if a scheme is
convergent overall and its nite-difference analogs are convergent and consistent.
Other problems arising from nite-difference and other solutions to partial
differential equations are numerical diffusion (articial spreading of peak values
across several grid cells) and numerical dispersion (waves appearing ahead of and
behind peak values). These problems can usually be mitigated by increasing the
resolution of the spatial grid (e.g., decreasing Lx), decreasing the time step, or
increasing the order of approximation of the nite-difference analog.
6.4.2 Low-order approximations of derivatives
A nite-difference approximation of a differential, such as ux, involves the
replacement of individual differential expressions, such as du or dx, with nite-
difference analogs, such as Lu or Lx, respectively. Suppose ux is discretized
over a westeast grid, as shown in Fig. 6.2, where all grid cells are rectangular.
Each cell is denoted by an index number i, and the distance from the western edge
of the entire grid to the western edge of cell i is x
i
.
On the grid layout just dened, the differential scalar velocity du at point x
i
can
be approximated as Lu
i
= u
i +1
u
i 1
, Lu
i
= u
i +1
u
i
, or Lu
i
= u
i
u
i 1
, which
are the central-, forward-, and backward-difference approximations, respectively.
The corresponding discretizations of dx are Lx
i
= x
i +1
x
i 1
, Lx
i
= x
i +1
x
i
,
and Lx
i
= x
i
x
i 1
, respectively. In the central-difference case, the slope of the
tangent at point x
i
in Fig. 6.2 is approximately
u
x

Lu
i
Lx
i
=
u
i +1
u
i 1
x
i +1
x
i 1
(6.10)
Similar equations can be written for the forward- and backward-difference cases.
The approximations just discussed can be derived from a Taylor series expan-
sion. If gas concentration is a continuous function of westeast distance, as shown
in Fig. 6.3, the values of N at points x +Lx and x Lx, respectively, are deter-
mined from Taylors theorem as
N
x+Lx
= N
x
+Lx
N
x
x
+
1
2
Lx
2
2
N
x
x
2
+
1
6
Lx
3
3
N
x
x
3
+
1
24
Lx
4
4
N
x
x
4
+
(6.11)
N
xLx
= N
x
Lx
N
x
x
+
1
2
Lx
2
2
N
x
x
2

1
6
Lx
3
3
N
x
x
3
+
1
24
Lx
4
4
N
x
x
4

(6.12)
If grid spacing is uniform (Lx is constant), the sum of (6.11) and (6.12) is
N
x+Lx
+ N
xLx
= 2N
x
+Lx
2
2
N
x
x
2
+
1
12
Lx
4
4
N
x
x
4
+ (6.13)
Rearranging (6.13) gives
2
N
x
x
2
=
N
x+Lx
2N
x
+ N
xLx
Lx
2
+ O(Lx
2
) (6.14)
175
N
x
N
x
x
i1
x
i
x
i+1
A
B
C
xx x x+x
x x
Figure 6.3 Derivative approxima-
tions at a point on a continuous
function. The derivative at point B is
approximated with chords AC, BC,
or AB, which give the slope of the
tangent at point B for the central-,
forward-, and backward-difference
approximations, respectively.
where
O(Lx
2
) =
1
12
Lx
2
4
N
x
x
4
(6.15)
includes all terms of order Lx
2
and higher. If O(Lx
2
) is small, (6.14) simplies to
2
N
x
x
2

N
x+Lx
2N
x
+ N
xLx
Lx
2
(6.16)
where O(Lx
2
) is now the truncation error. Equation (6.16) is a second-order
central-difference approximation of
2
N
x
x
2
. The equation is second-order
because the lowest-order exponent in the truncation error is two. It is a central-
difference approximation because it relies on equally weighted values of N on each
side of node x.
Subtracting (6.12) from (6.11) gives
N
x+Lx
N
xLx
= 2Lx
N
x
x
+
1
3
Lx
3
3
N
x
x
3
+ (6.17)
Rearranging this equation results in
N
x
x
=
N
x+Lx
N
xLx
2Lx
+ O(Lx
2
) (6.18)
where
O(Lx
2
) =
1
6
Lx
2
3
N
x
x
3
(6.19)
176
includes all terms of order Lx
2
and higher. If O(Lx
2
) is small, (6.18) simplies
to
N
x
x

N
x+Lx
N
xLx
2Lx
=
N
i +1
N
i 1
2Lx
(6.20)
where i + 1 and i 1 are surrogates for x +Lx and x Lx, respectively. This
equation is a second-order central-difference approximation of the rst derivative
of N
x
. Equation (6.20) gives the slope of the tangent (represented by chord AC) of
N
x
at point B in Fig. 6.3.
Another approximation of the rst derivative of N
x
is obtained from the rst
two terms of (6.11). Rearranging these terms gives
N
x
x

N
x+Lx
N
x
Lx
=
N
i +1
N
i
Lx
(6.21)
The truncated portion of the approximation includes terms rst-order and higher
[O(Lx)]; thus, (6.21) is a rst-order forward-difference approximation of the rst
derivative of N
x
. The slope of this derivative is represented by chord BC in Fig. 6.3.
Rearranging the rst two terms of (6.12) gives
N
x
x

N
x
N
xLx
Lx
=
N
i
N
i 1
Lx
(6.22)
which is the rst-order backward-difference approximation of the rst derivative
of N
x
, represented by chord AB in Fig. 6.3.
If time, not space, is the independent variable, the second-order central-, rst-
order forward-, and rst-order backward-difference approximations of N
t
t
are
N
t
t

N
t+h
N
th
2h
N
t
t

N
t+h
N
t
h
N
t
t

N
t
N
th
h
(6.23)
respectively, where h = Lt is the time-step size, t is the current time, t +h is one
time step forward, and t h is one time step backward. These equations are derived
in the same manner as (6.20), (6.21), and (6.22), respectively.
6.4.3 Arbitrary-order approximations of derivatives
Finite-difference approximations of arbitrary order can be obtained systematically
(e.g., Celia and Gray 1992). The approximation of
m
Nx
m
, which is the mth
derivative of N, can be obtained by expanding the derivative across q discrete
nodes in the x-direction. If the independent variable is time, the derivative can
be expanded along q time steps. The minimum number of nodes allowed in the
177
x
x
1
x
2
x
3
x
4
x
5
*
Figure 6.4 Grid spacing of an arbitrary-spaced grid where q =5. The deriva-
tive is taken at node point x
3
, marked
.
expansion is m + 1. In general, the maximum order of approximation of a nite-
difference solution is q m, although it may be smaller or larger for some individual
cases. For instance, when m is even and the grid spacing is constant, the order of
approximation can by increased to q m + 1.
Figure 6.4 shows the arbitrary grid spacing for the derivation to come. The
location at which the derivative is taken does not need to correspond to a node
point, although in the gure the derivative is assumed to be taken at node point
x
3
. The distance between two node points is Lx
i
= x
i +1
x
i
, where i varies from
1 to q 1.
The nite-difference solution to the mth derivative across q nodes is approxi-
mately
m
N
x
m

q
i =1
i
N
i
=
1
N
1
+
2
N
2
+ +
q
N
q
(6.24)
where the
i
s are constants to be determined. A Taylor series expansion of N at
node i across the point at which the derivative is taken (
) is
N
i
= N
+(x
i
x
)
N
x
+
1
2
(x
i
x
)
2
2
N
x
2
+
1
6
(x
i
x
)
3
3
N
x
3
+ (6.25)
Combining (6.24) with (6.25) and gathering terms gives
m
N
x
m

q
i =1
i
N
i
=
q
i =1
i
N
+
q
i =1
i
(x
i
x
)
N
x
+
q
i =1
i
1
2
(x
i
x
)
2
2
N
x
2
+
(6.26)
This equation can be rewritten as
q
i =1
i
N
i
= B
0
N
+ B
1
N
x
+ B
2
2
N
x
2
+ (6.27)
where
B
n
=
q
i =1
i
1
n!
(x
i
x
)
n
for n = 0 q 1 (6.28)
178
Equation (6.28) represents a matrix of q equations and unknowns. Multiplying
(6.28) by n! gives
1 1 1 1
(x
1
x
) (x
2
x
) (x
3
x
) (x
q
x
)
(x
1
x
)
2
(x
2
x
)
2
(x
3
x
)
2
(x
q
x
)
2
.
.
.
.
.
.
.
.
.
.
.
.
(x
1
x
)
q1
(x
2
x
)
q1
(x
3
x
)
q1
(x
q
x
)
q1
3
.
.
.
0!B
0
1!B
1
2!B
2
.
.
.
(q 1)!B
q1
(6.29)
The highest-order derivative is found when B
n
= 1 for n = m and B
n
= 0 for all
other n.
The rst-order backward-difference approximation of Nx(m = 1) is found
from (6.29) by discretizing Nx across two equally spaced grid cells (q = 2),
setting B
1
= 1, and setting B
n
= 0 for all other n. The resulting matrix is
1 1
Lx 0

i 1
0
1
(6.30)
where the subscript i 1 indicates one node to the left of i. The matrix has solution
i 1
= 1Lx and
i
= 1Lx. Substituting these coefcients into
N
x

1
N
1
+
2
N
2
=
i 1
N
i 1
+
i
N
i
(6.31)
from (6.24) gives the approximation shown in (6.22) and Table 6.2 (a).
Second-order central- and backward-difference approximations of Nx
(m = 1) are found by discretizing Nx across three nodes (q = 3). The resulting
matrices are
1 1 1
Lx 0 Lx
(Lx)
2
0 (Lx)
2
i 1
i +1
0
1
0
1 1 1
2Lx Lx 0
(2Lx)
2
(Lx)
2
0
i 2
i 1
0
1
0
(6.32)
respectively. Substituting solutions to these matrices into (6.24) gives the approx-
imation shown in Table 6.2 (c) and (d), respectively. The second-order forward-
difference approximation of Nx is found by discretizing around the rst column
in (6.32). The result is shown Table 6.2 (e). Forward- and backward-difference dis-
cretizations are negatively symmetric to each other.
Third-order backward- and forward-difference approximations of Nx are
found in a similar manner. The results are shown in Table 6.2 (f) and (g), respec-
tively, where the discretizations are around four cells.
179
Table 6.2 Finite-difference approximations of Nx and
2
Nx
2
Order m q Approximation
(a) First-order backward 1 2
N
x

N
i
N
i 1
Lx
(b) First-order forward 1 2
N
x

N
i +1
N
i
Lx
(c) Second-order central 1 3
N
x

N
i +1
N
i 1
2Lx
(d) Second-order backward 1 3
N
x

N
i 2
4N
i 1
+3N
i
2Lx
(e) Second-order forward 1 3
N
x

3N
i
+4N
i +1
N
i +2
2Lx
(f) Third-order backward 1 4
N
x

N
i 2
6N
i 1
+3N
i
+2N
i +1
6Lx
(g) Third-order forward 1 4
N
x

2N
i 1
3N
i
+6N
i +1
N
i +2
6Lx
(h) Fourth-order central 1 5
N
x

N
i 2
8N
i 1
+8N
i +1
N
i +2
12Lx
(i) Fourth-order backward (I) 1 5
N
x

N
i 3
+6N
i 2
18N
i 1
+10N
i
+3N
i +1
12Lx
(j) Fourth-order forward (I) 1 5
N
x

3N
i 1
10N
i
+18N
i +1
6N
i +2
+ N
i +3
12Lx
(k) Fourth-order backward (II) 1 5
N
x

3N
i 4
+16N
i 3
36N
i 2
+48N
i 1
25N
i
12Lx
(l) Fourth-order forward (II) 1 5
N
x

25N
i
48N
i +1
+36N
i +2
16N
i +3
+3N
i +4
12Lx
(m) Second-order central 2 3

2
N
x
2

N
i +1
2N
i
+ N
i 1
Lx
2
(n) Fourth-order central 2 5

2
N
x
2

N
i 2
+16N
i 1
30N
i
+16N
i +1
N
i +2
12Lx
2
A fourth-order central-difference approximation of Nx is found from
1 1 1 1 1
2Lx Lx 0 Lx 2Lx
(2Lx)
2
(Lx)
2
0 (Lx)
2
(2Lx)
2
(2Lx)
3
(Lx)
3
0 (Lx)
3
(2Lx)
3
(2Lx)
4
(Lx)
4
0 (Lx)
4
(2Lx)
4
i 2
i 1
i +1
i +2
0
1
0
0
0
(6.33)
The approximation resulting fromthis matrix is shown in Table 6.2 (h). One fourth-
order backward-difference approximation of Nx is obtained by solving (6.33)
after discretizing around the fourth instead of the third column in the equation. The
result is shown in Table 6.2 (i). The corresponding fourth-order forward-difference
approximation is shown in Table 6.2 (j). Another fourth-order backward-
difference approximation of Nx is obtained by solving (6.33) after discretiz-
ing around the fth column in the equation. The result appears in Table 6.2 (k).
The corresponding fourth-order forward-difference approximation appears in
180
Table 6.2 (l). A fourth-order central-difference approximation of
2
Nx
2
is
obtained by solving (6.33), but setting B
2
= 1 and B
n
= 0 for all other n. The
solution is shown in Table 6.2 (n).
6.4.4 Time-stepping schemes for the advectiondiffusion equation
Finite-difference approximations can be applied to the temporal and spatial deriva-
tives of the advectiondiffusion equations given in (6.1)(6.3). In the following
subsections, approximations to temporal derivatives are discussed with respect to
the westeast form of the advectiondiffusion equation.
6.4.4.1 CourantFriedrichsLewy stability criterion
Some time-stepping schemes are explicit, whereas others are implicit or semiim-
plicit. In an explicit time-stepping scheme, nal terms (time t) are calculated explic-
itly from known values (e.g., from values at times t h, t 2h, etc.). In an implicit
time-stepping scheme, nal terms (time t) are evaluated from other terms at time t,
which are initially unknown but which are solved simultaneously with the desired
terms. In a semiimplicit scheme, nal terms are evaluated from some terms that are
known (times t h, t 2h, etc.) and other terms that are unknown (time t).
Whereas explicit and semiimplicit time-stepping schemes are generally condi-
tionally stable (stable for any time step below a specied value), some implicit
schemes may be unconditionally stable (stable regardless of time step). Such
schemes, by nature, do not require iteration. Other implicit schemes require itera-
tion, and these are conditionally stable.
The time step limitation for an explicit solution to the advection equation
can be approximated with the CourantFriedrichsLewy (CFL) stability criterion
(Courant et al. 1928). The advection equation in the westeast direction, for exam-
ple, is found by removing the diffusion termfrom(6.1). When the resulting equation
is solved explicitly, stability is generally maintained when the CFL criterion,
h - Lx
min
|u
max
| (6.34)
is met, where |u
max
| is the maximum westeast wind speed, and Lx
min
is the min-
imum westeast grid-cell length in the domain. If the maximum wind speed is
|u
max
| =20 m s
1
, for example, and the minimum grid-cell length is Lx
min
=5 km,
the CFL criterion predicts that the maximum time step for maintaining stability is
h = 250 s. To maintain stability, a parcel of air is not allowed to travel across a
grid cell during a single time step.
For equations more general than the advectiondiffusion equation, |u
max
| should
be replaced by |c
max
|, where |c
max
| is the maximum speed of propagation in the
domain. In the case of the primitive equations, the maximum speed of propaga-
tion is the speed of horizontally propagating acoustic waves (the speed of sound)
(Section 5.1).
If advection is ignored, (6.1) simplies to the westeast diffusion equation. A
stability criterion for this equation, analogous to (6.34), is h - Lx
2
min
K
max
, where
181
K
max
is the largest eddy diffusion coefcient in the domain. For a typical vertical
eddy diffusion coefcient of 50 m
2
s
1
, for example, this stability criterion suggests
that the time step for eddy diffusion through an altitude of 100 m needs to be less
than 200 s. For a typical horizontal eddy diffusion coefcient of 2500 m
2
s
1
, the
time step for eddy diffusion across a 5-km cell must be less than 10 000 s. Explicit
solutions to the diffusion equation are more likely to become unstable in the vertical
than in the horizontal in an atmospheric model.
6.4.4.2 Forward Euler scheme
A basic time-discretization scheme for the advectiondiffusion equation is the for-
ward Euler scheme. If u varies and K = K
h.xx
is constant along x, and if u and K
are constant during a time step, the time, advection, and diffusion derivatives in
(6.1) can be discretized with (6.23), (6.20), and (6.16), respectively, to yield
N
i.t
N
i.th
h
+
(uN)
i +1.th
(uN)
i 1.th
2Lx
K
N
i +1.th
2N
i.th
+ N
i 1.th
Lx
2
= 0
(6.35)
The temporal and spatial derivatives in this equation are rst- and second-order
approximations, respectively. The equation is called the forward-in-time, centered-
in-space (FTCS) approximation because the time derivative uses information from
one previous time step, and the advection terms are central-difference expressions.
Since all terms in (6.35), except the nal concentration, are evaluated at time t h,
the FTCS approximation is explicit. When a time derivative is rst-order and spatial
derivatives are determined explicitly, as in the example above, the time scheme is
a forward Euler one. For all values of u, this equation is unconditionally unstable
for K = 0 and for large values of K and conditionally stable for small values of K,
except when K = 0 (Mesinger and Arakawa 1976).
Because values on the right side of (6.35) are known, the equation can be solved
immediately for i = 1, . . . , I (where I is the number of westeast nodes). If
the grid contains lateral boundaries, the solution depends on the terms (uN)
0.th
,
(uN)
I+1.th
, (KN)
0.th
, and (KN)
I+1.th
, which lie beyond the boundaries. Outside
boundary values may be set equal to values just inside the boundary (e.g., at nodes
i = 1 and i = I) from the previous time step. When the grid has periodic boundary
conditions (e.g., a grid that has no lateral boundaries because it wraps around on
itself), node i = 0 is also node i = I, and node i = 1 is also node i = I + 1. In such
a case, outside boundary values are not needed.
6.4.4.3 Implicit scheme
Equation (6.35) can be solved implicitly by evaluating all terms on the right side
at time t. In implicit form, (6.35) becomes
N
i.t
N
i.th
h
+
(uN)
i +1.t
(uN)
i 1.t
2Lx
K
N
i +1.t
2N
i.t
+ N
i 1.t
Lx
2
= 0 (6.36)
182
The solution to (6.36) along i = 1, . . . , I is obtained by rearranging the equation
as
A
i
N
i 1.t
+ B
i
N
i.t
+ D
i
N
i +1.t
= N
i.th
(6.37)
where
A
i
= h
u
2Lx
+
K
Lx
2
i 1
B
i
= 1 +h
2K
Lx
2
i
D
i
= h
u
2Lx

K
Lx
2
i +1
(6.38)
For a limited-area domain (a domain with boundaries at both ends), the matrix
arising from (6.37) is
B
1
D
1
0 0 0 0 0
A
2
B
2
D
2
0 0 0 0
0 A
3
B
3
D
3
0 0 0
0 0 A
4
B
4
0 0 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 0 0 B
I2
D
I2
0
0 0 0 0 A
I1
B
I1
D
I1
0 0 0 0 0 A
I
B
I
N
1.t
N
2.t
N
3.t
N
4.t
.
.
.
N
I2.t
N
I1.t
N
I.t
N
1.th
N
2.th
N
3.th
N
4.th
.
.
.
N
I2.th
N
I1.th
N
I.th
A
1
N
0.t
0
0
0
.
.
.
0
0
D
I
N
I+1.t
(6.39)
where A
1
N
0.t
and D
I
N
I+1.t
are outside boundary values. Outside values are
assumed to be known in advance although they carry the subscript t. Equation
(6.39) is a tridiagonal matrix, which is solved by matrix decomposition and back-
substitution in the order,
Decomposition
1
=
D
1
B
1
i
=
D
i
B
i
+ A
i
i 1
for i = 2. . . . . I
1
=
R
1
B
1
i
=
R
i
A
i
i 1
B
i
+ A
i
i 1
for i = 2. . . . . I (6.40)
Backsubstitution
N
I.t
=
I
N
i.t
=
i
+
i
N
i +1.t
for i = I 1. . . . . 1. 1 (6.41)
where R
i
represents the right side of (6.39). The solution to (6.39) is mass con-
serving and unconditionally stable for all values of u and K, but it is numerically
diffusive. To obtain the solution, (6.36) was converted from a partial differential
equation (dependent on time and space) to an ordinary differential equation (depen-
dent on time only). The linear ODE was then solved by matrix decomposition and
backsubstitution.
183
For a domain with periodic boundary conditions (Section 6.4.4.2), the matrix
for (6.39) becomes
B
1
D
1
0 0 0 0 A
1
A
2
B
2
D
2
0 0 0 0
0 A
3
B
3
D
3
0 0 0
0 0 A
4
B
4
0 0 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 0 0 B
I2
D
I2
0
0 0 0 0 A
I1
B
I1
D
I1
D
I
0 0 0 0 A
I
B
I
N
1.t
N
2.t
N
3.t
N
4.t
.
.
.
N
I2.t
N
I1.t
N
I.t
N
1.th
N
2.th
N
3.th
N
4.th
.
.
.
N
I2.th
N
I1.th
N
I.th
(6.42)
where values at node I are adjacent to those at node 1. The solution is obtained by
solving (6.40), followed by
1
=
A
1
B
1
i
=
A
i
i 1
B
i
+ A
i
i 1
for i = 2. . . . . I
I
= 1
i
=
i
i +1
+
i
for i = I 1. . . . . 1. 1
I
= 0
i
=
i
i +1
+
i
for i = I 1. . . . . 1. 1
N
I.t
=
I

1
D
I
B
I
+ A
I
I1
1 +
D
I
1
+ A
I
I
B
I
+ A
I
I1
N
i.t
=
i
+
i
N
i +1.t
+
i
N
I.t
for i = I 1. . . . . 1. 1 (6.43)
This solution is mass-conserving but does not require outside boundary informa-
tion.
6.4.4.4 CrankNicolson scheme
The implicit approximation just described was rst order in time and second order
in space. The order of approximation in time can be improved to second order
with the CrankNicolson (trapezoidal) scheme (Crank and Nicolson 1947). This
scheme is semiimplicit since some terms on the right side are evaluated at time t
and others are evaluated at time t h. With the CrankNicolson scheme, spatial
derivatives are weighted 50 percent between the initial and nal times. Rewriting
(6.35) gives
N
i.t
N
i.th
h
+
j
c
(uN)
i +1.t
(uN)
i 1.t
2Lx
+(1 j
c
)
(uN)
i +1.th
(uN)
i 1.th
2Lx
j
c
N
i +1.t
2N
i.t
+ N
i 1.t
Lx
2
+(1 j
c
)
N
i +1.th
2N
i.th
+ N
i 1.th
Lx
2
= 0
(6.44)
where j
c
is the CrankNicolson parameter. When a nite-difference equation is
written in terms of j
c
, the equation is in CrankNicolson form. When j
c
= 0.5,
184
(6.44) reduces to the CrankNicolson scheme, which is second order in time and
unconditionally stable for all values of u and K. When j
c
= 0, (6.44) reduces to
(6.35), the forward Euler scheme, and when j
c
= 1, (6.44) reduces to (6.36), the
implicit scheme. Equation (6.44) can be rewritten as
A
i
N
i 1.t
+ B
i
N
i.t
+ D
i
N
i +1.t
= E
i
N
i 1.th
+ F
i
N
i.th
+ G
i
N
i +1.th
(6.45)
where
A
i
= j
c
h
u
2Lx
+
K
Lx
2
i 1
B
i
= 1 +j
c
h
2K
Lx
2
i
D
i
= j
c
h
u
2Lx

K
Lx
2
i +1
E
i
= (1 j
c
) h
u
2Lx
+
K
Lx
2
i 1
F
i
= 1 (1 j
c
) h
2K
Lx
2
i
G
i
= (1 j
c
) h
u
2Lx

K
Lx
2
i +1
(6.46)
When a grid has lateral boundaries, the matrix arising from (6.45) is the same as
(6.39), except that the right side of (6.39) is replaced with
=
F
1
G
1
0 0 0 0 0
E
2
F
2
G
2
0 0 0 0
0 E
3
F
3
G
3
0 0 0
0 0 E
4
F
4
0 0 0
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
0 0 0 0 F
I2
G
I2
0
0 0 0 0 E
I1
F
I1
G
I1
0 0 0 0 0 E
I
F
I
N
1.th
N
2.th
N
3.th
N
4.th
.
.
.
N
I2.th
N
I1.th
N
I.th
E
1
N
0.th
A
1
N
0.t
0
0
0
.
.
.
0
0
G
I
N
I+1.th
D
I
N
I+1.t
(6.47)
Equations (6.39) and (6.47) are solved with (6.40) and (6.41). For a domain with
periodic boundary conditions, (6.39) and (6.47) are solved after the rightmost
column in (6.47) is removed, after A
1
and D
I
are placed in the top right and
bottom left corners, respectively, of (6.39), and after E
1
and G
I
are placed in the
top right and bottom left corners, respectively, of (6.47).
6.4.4.5 Leapfrog scheme
Another scheme that increases the order of approximation in time to second order
is the leapfrog scheme. This scheme uses information from two previous time steps
to predict information for a third. More specically, spatial derivatives from time
t h are used to evaluate differences between times t and t 2h. The leapfrog
solution to the westeast advectiondiffusion equation is
N
i.t
N
i.t2h
2h
+
(uN)
i +1.th
(uN)
i 1.th
2Lx
K
N
i +1.th
2N
i.th
+ N
i 1.th
Lx
2
= 0
(6.48)
where values from time t h and t 2h are determined from previous time steps.
185
The discretization above is second-order in time and space. When used alone,
the leapfrog scheme is unconditionally unstable for all nonzero values of K. When
K = 0, the leapfrog scheme is conditionally stable for linear equations. For nonlin-
ear equations, the scheme destabilizes over time. To suppress such instability, com-
putations fromanother scheme must be inserted every fewleapfrog steps (Mesinger
and Arakawa 1976). Ascheme used to stabilize the leapfrog scheme is the Matsuno
scheme.
6.4.4.6 Matsuno scheme
The Matsuno scheme (Matsuno 1966) is an explicit time-stepping scheme com-
monly used to stabilize and initialize the leapfrog scheme. With the Matsuno
scheme, time derivatives are estimated with a forward-difference approximation.
The estimated values (subscript est) are substituted into the spatial derivatives
to estimate nal values. The estimation and correction steps are
N
i.est
N
i.th
h
+
(uN)
i +1.th
(uN)
i 1.th
2Lx
K
N
i +1.th
2N
i.th
+ N
i 1.th
Lx
2
= 0
(6.49)
N
i.t
N
i.th
h
+
(uN)
i +1.est
(uN)
i 1.est
2Lx
K
N
i +1.est
2N
i.est
+ N
i 1.est
Lx
2
= 0
(6.50)
respectively. Although the Matsuno scheme requires twice as many computations
as does either the forward Euler or leapfrog scheme per time step, the Matsuno
scheme is still a rst-order approximation in time. It is conditionally stable for all
values of u when K is zero or small but absolutely unstable for large values of K
(Mesinger and Arakawa 1976). When combined with the leapfrog scheme to solve
the equations of atmospheric dynamics, Matsuno steps are usually taken every
515 leapfrog steps.
6.4.4.7 Heun scheme
With the Heun scheme, time derivatives are estimated with a forward-difference
approximation that uses initial values in the spatial derivative. Final time deriva-
tives are determined with an average spatial derivative. The average is taken as
one-half the spatial derivative determined from initial values plus one-half the spa-
tial derivative determined from estimated values. With respect to the advection
diffusion equation, the Heun scheme involves solving (6.49) followed by
N
i.t
N
i.th
h
+
1
2
(uN)
i +1.est
(uN)
i 1.est
2Lx

K
2
N
i +1.est
2N
i.est
+ N
i 1.est
Lx
2
+
1
2
(uN)
i +1.th
(uN)
i 1.th
2Lx

K
2
N
i +1.th
2N
i.th
+ N
i 1.th
Lx
2
= 0 (6.51)
The Heun scheme is a second-order approximation in time. For all values of
u, this scheme is unconditionally unstable when K = 0 and when K is large,
186
and conditionally stable when K is small and nonzero (Mesinger and Arakawa
1976).
6.4.4.8 AdamsBashforth scheme
Another time-differencing scheme is a simplied version of the AdamsBashforth
scheme. Like the leapfrog scheme, this scheme is explicit, uses three time levels, and
is a second-order approximation in time. The scheme discretizes the advection
diffusion equation as
N
i.t
N
i.th
h
+
3
2
(uN)
i +1.th
(uN)
i 1.th
2Lx

3
2
K
N
i +1.th
2N
i.th
+ N
i 1.th
Lx
2
1
2
(uN)
i +1.t2h
(uN)
i 1.t2h
2Lx
+
1
2
K
N
i +1.t2h
2N
i.t2h
+ N
i 1.t2h
Lx
2
= 0
(6.52)
where the t h time level is favored over the t 2h time level. For all values of u,
the AdamsBashforth scheme is unconditionally unstable when K = 0 and when
K is large, and conditionally stable when K is small and nonzero (Mesinger and
Arakawa 1976). This scheme is useful for short integration periods when a small
time step is taken.
6.4.4.9 Fourth-order RungeKutta scheme
The last time-differencing scheme discussed is the fourth-order RungeKutta
scheme (e.g., Press et al. 1992). This scheme is explicit and requires information
from one time step backward only, but makes three guesses before forecasting
nal values for the time step. When the RungeKutta scheme is applied to the
advectiondiffusion equation, the concentration at time t is calculated with
N
i.t
= N
i.th
+
k
1
6
+
k
2
3
+
k
3
3
+
k
4
6
(6.53)
where
k
1
= h
(uN)
i +1.th
(uN)
i 1.th
2Lx
+ K
N
i +1.th
2N
i.th
+ N
i 1.th
Lx
2
k
2
= h
(u
th
N
est1
)
i +1
(u
th
N
est1
)
i 1
2Lx
+ K
N
i +1.est1
2N
i.est1
+ N
i 1.est1
Lx
2
k
3
= h
(u
th
N
est2
)
i +1
(u
th
N
est2
)
i 1
2Lx
+ K
N
i +1.est2
2N
i.est2
+ N
i 1.est2
Lx
2
k
4
= h
(u
th
N
est3
)
i +1
(u
th
N
est3
)
i 1
2Lx
+ K
N
i +1.est3
2N
i.est3
+ N
i 1.est3
Lx
2
(6.54)
and
N
i.est1
= N
i.th
+
k
1
2
N
i.est2
= N
i.th
+
k
2
2
N
i.est3
= N
i.th
+k
3
(6.55)
187
10
11
10
9
10
7
10
5
10
3
10
1
0.1 1 10
E
r
r
o
r
Time step (s)
1
st
-order
5
th
6
th
-order
2
nd
-order
Figure 6.5 Comparison of convergence
among four schemes applied to the one-
dimensional advectiondiffusion equation for
a tracer concentration. RungeKutta,
----- AdamsBashforth, Matsuno,
forward Euler. The wind speed and diffusion
coefcient were constant at 5 ms
1
and
1000 m
2
s
1
, respectively. Boundary condi-
tions were periodic. The actual orders of
approximation determined by the curves are
shown. From Ketean and Jacobson (2005a).
When K = 0, the scheme is stable when h 2
2Lx|u
max
|, where |u
max
| is the
maximum westeast wind speed. When |u
max
| = 0, the scheme is stable when
h CLx
2
K
max
, where K
max
is the maximum diffusion coefcient on the grid and
C is a constant between 0.25 and 0.3. When K > 0 and |u
max
| > 0, the stability of
the scheme is a complex function of |u
max
|, K
max
, and grid spacing (Ketean 2005;
Ketean and Jacobson 2005a).
Figure 6.5 compares the convergence of the fourth-order RungeKutta, Adams
Bashforth, Matsuno, and forward Euler schemes. The Matsuno and forward Euler
schemes are rst-order approximations in time, as illustrated in the gure, which
shows that when the time step is reduced by half, errors of the schemes decrease
by half. The AdamsBashforth scheme is a second-order approximation in time.
Thus, a factor of two reduction in the time step reduces its error by a factor of
four. A factor of two reduction in the fourth-order RungeKutta time step ideally
reduces its error by a factor of 16. The gure, though, shows that a reduction in
the time step by a factor of two reduced the error by a factor of 3641, making
this scheme a fthsixth order scheme for this application. The reason is that the
order of approximation is the order of the lowest-order term in a Taylor series
expansion neglected in the approximation. In the present case, the lowest-order
term neglected was negligible, thus the order of approximation was governed by
the second-lowest-order term neglected.
Figure 6.5 shows that not only does a high-order scheme decrease the error
with decreasing time step to a greater extent than does a low-order scheme, but a
high-order scheme also results in a lower error at any given time step than does a
low-order scheme.
188
6.4.5 Fourth-order in space solution to the
advectiondiffusion equation
The time-difference schemes discussed above can be applied with higher-order
spatial nite-difference approximations. For example, substituting fourth-order
expansions of the advection and diffusion terms from Table 6.2 (h) and (n), respec-
tively, into (6.1) gives a fully implicit, unconditionally stable formof the advection
diffusion equation as
N
i.t
N
i.th
h
+
(uN)
i 2.t
8(uN)
i 1.t
+8(uN)
i +1.t
(uN)
i +2.t
12Lx
K
N
i 2.t
+16N
i 1.t
30N
i.t
+16N
i +1.t
N
i +2.t
12Lx
2
= 0 (6.56)
This equation can be written in banded-matrix form with ve terms on both sides
and solved with a banded matrix method (e.g., Press et al. 1992). Such methods
may combine matrix decomposition and backsubstitution with a sparse-matrix
technique to reduce the number of computations.
6.4.5.1 Variable grid spacing and eddy diffusion coefcients
The previous solutions to the advectiondiffusion equation were obtained by
assuming constant grid spacing and eddy diffusion coefcients. Here a solution
that assumes variable grid spacing and diffusion coefcients is considered.
The westeast advection term in (6.1) can be discretized with (6.24) as
(uN)
x
=
a.i 1
(uN)
i 1
+
a.i
(uN)
i
+
a.i +1
(uN)
i +1
(6.57)
When grid spacing is variable, the second-order central-difference approximation
coefcients for this equation are obtained by solving the matrix equation from
(6.29) with B
1
= 1 and all other B
n
= 0,
1 1 1
(x
i
x
i 1
) 0 (x
i +1
x
i
)
(x
i
x
i 1
)
2
0 (x
i +1
x
i
)
2
a.i 1
a.i
a.i +1
0
1
0
(6.58)
to yield
a.i 1
=
(x
i +1
x
i
)
(x
i
x
i 1
) (x
i +1
x
i 1
)
a.i
=
(x
i +1
x
i
) (x
i
x
i 1
)
(x
i +1
x
i
) (x
i
x
i 1
)
(6.59)
a.i +1
=
x
i
x
i 1
(x
i +1
x
i
) (x
i +1
x
i 1
)
189
The westeast diffusion term in (6.1) expands to
K
N
x
=
K
x
N
x
+ K
2
N
x
2
(6.60)
The second-order central-difference approximations to the terms on the right side
of this equation are
K
x

a.i 1
K
i 1
+
a.i
K
i
+
a.i +1
K
i +1
N
x

a.i 1
N
i 1
+
a.i
N
i
+
a.i +1
N
i +1
(6.61)
K
2
N
x
2
K
i
(
d.i 1
N
i 1
+
d.i
N
i
+
d.i +1
N
i +1
)
where the
a
terms are the same as those in (6.59). The
d
terms are found by
solving the matrix equation for a second-order central-difference approximation
from (6.29) with B
2
= 1 and all other B
n
= 0,
1 1 1
(x
i
x
i 1
) 0 (x
i +1
x
i
)
(x
i
x
i 1
)
2
0 (x
i +1
x
i
)
2
d.i 1
d.i
d.i +1
0
0
2
(6.62)
to yield
d.i 1
=
2
(x
i
x
i 1
) (x
i +1
x
i 1
)
d.i
=
2
(x
i +1
x
i
) (x
i
x
i 1
)
(6.63)
d.i +1
=
2
(x
i +1
x
i
) (x
i +1
x
i 1
)
K
N
x

K.i 1
N
i 1
+
K.i
N
i
+
K.i +1
N
i +1
(6.64)
where
K.i 1
= (
a.i 1
K
i 1
+
a.i
K
i
+
a.i +1
K
i +1
)
a.i 1
+ K
i
d.i 1
K.i
= (
a.i 1
K
i 1
+
a.i
K
i
+
a.i +1
K
i +1
)
a.i
+ K
i
d.i
(6.65)
K.i +1
= (
a.i 1
K
i 1
+
a.i
K
i
+
a.i +1
K
i +1
)
a.i +1
+ K
i
d.i +1
Applying (6.57) and (6.65) to (6.1) gives a second order in space solution to the
advectiondiffusion equation that allows for variable grid spacing, wind speeds,
190
and eddy diffusion coefcients. In CrankNicolson form, the advectiondiffusion
equation is
N
i.t
N
i.th
h
= j
c
{[(
a
u
K
)N]
i 1
+[(
a
u
K
)N]
i
+[(
a
u
K
)N]
i +1
}
t
(1 j
c
){[(
a
u
K
)N]
i 1
+[(
a
u
K
)N]
i
+[(
a
u
K
)N]
i +1
}
th
(6.66)
This equation can be written in tridiagonal matrix form and solved.
6.4.6 Finite-differencing in two directions
Combining the westeast and southnorth terms in the advectiondiffusion equa-
tion gives the horizontal form of the equation as
N
t
+
(uN)
x
+
(vN)
y

x
K
h.xx
N
x
K
h.yy
N
y
= 0 (6.67)
When constant grid spacing and eddy diffusion coefcients are assumed, the
implicit expansion of (6.67) to second order in space is
N
i. j.t
N
i. j.th
h
+
(uN)
i +1. j
(uN)
i 1. j
2Lx
+
(vN)
i. j +1
(vN)
i. j 1
2Ly
K
h.xx
N
i 1. j
2N
i. j
+ N
i +1. j
Lx
2
+ K
h.yy
N
i. j 1
2N
i. j
+ N
i. j +1
Ly
2
t
= 0 (6.68)
where j is the grid index in the northsouth direction. This equation is linear for all
i and j and can be solved implicitly in matrix form, just as with the one-dimensional
case. Although the matrix is not banded, it may be solved by decomposition and
backsubstitution. Equation (6.66) can also be extended to three dimensions, put
in CrankNicolson form, and solved with variable grid spacing and eddy diffusion
coefcients.
The primary disadvantage of solving (6.68) implicitly is that the matrix order
quickly becomes large. For a 100 100 horizontal grid, the required matrix is
10 000 10 000. The advantage of solving the equation implicitly is that the
solution is unconditionally stable for all values of u, v, K
h.xx
, and K
h.yy
. If (6.68)
is solved explicitly, it is unconditionally unstable for all values of u and v when
K = 0 or K is large and conditionally stable for all values of u and v for other
values of K.
6.4.7 The semi-Lagrangian method
A special case of the nite-difference method is the semi-Lagrangian method (e.g.,
Pepper et al. 1979; Robert 1982; Staniforth and Cote 1991; Makar and Karpik
1996; Yabe et al. 2001; Bermejo and Conde 2002; Nair et al. 2002). With this
method, the value of a variable at a specic location and time is obtained by rst
191
tracing back where the air parcel containing the variable came from during the
last time step. Suppose it is desired to nd the concentration of a gas at time t at
point B, located at the center of a model grid cell. If the wind speed near point
B is 5 m s
1
and the model time step is h = 300 s, the concentration at point B
and time t can be estimated as the concentration at time t h a distance Lx =
5 m s
1
300 s = 1500 m to the west of point B, dened now as point A. During
the time interval, the wind advects the gas from point A to B. Since point A is
not necessarily located at the center of a model grid cell, the gas concentration
at point A must be found by interpolating concentrations from adjacent grid-cell
centers.
In sum, with the semi-Lagrangian method, the concentration of a gas at time t
and location x is estimated as
N
x.t
= N
xuh.th
(6.69)
where N at location x uh and time t h is interpolated from values of N at
nearby node points. Mixing ratios, potential temperatures, and u- and v-scalar
velocities can be estimated in a similar manner. Equation (6.69) can be written for
one, two, or three directions.
Several methods exist to interpolate a variable to point A from nearby grid-cell
center values. One method is to estimate the slope of the variable between two
grid cells surrounding point A and to interpolate linearly between the two cells. A
more complex method is to t a polynomial through three or more adjacent cells
surrounding point A and to calculate the value at point A from the polynomial.
When the polynomial is a cubic, the method is the cubic spline method (Price and
MacPherson 1973; Purnell 1976).
An advantage of semi-Lagrangian schemes is that they can be run with a
long time step without concern for stability. Advection of tracers, such as water
vapor and potential temperature, is sometimes more accurate with a semi-
Lagrangian scheme than with a pure nite-difference scheme. A disadvantage of
semi-Lagrangian schemes is that, unless proper steps are taken, the mass of a trace
species is not conserved during advection. Transported mass can be conserved dur-
ing a time step if the tted spatial concentration curve, used for interpolation, is
normalized so that the integral of mass under the curve equals the total mass in the
system at the beginning of the time step.
6.5 SERIES EXPANSION METHODS
With a nite-difference method, each differential in a PDE is replaced with a dif-
ference analog written in terms of a nite number of values along a temporal or
spatial direction. With a series expansion method, a dependent variable (e.g., u, v,
w, N) in a PDE is replaced with a nite series that approximates its value. If the
192
6.5 Series expansion methods
PDE arising from the westeast advection equation at node i is
N
i
t
+
(uN)
i
x
= 0 (6.70)
a series-expansion approximation of the number concentration at node i is
N
i
N
i
(x) =
j
N
j
e
j
(x) (6.71)
where N
i
(x) is called a trial function. Assume for now that u is constant. The set
of j nodes over which the trial function is approximated is the trial space. The trial
function is the sum, over each node in the trial space, of the true concentration,
N
j
, multiplied by a basis function e
j
(x). The difference between (6.70) when N
i
(x)
is used and that when the summation over N
j
e
j
(x) is used is the residual, R
i
(x). A
residual is the difference between an approximate and an exact function.
A series expansion method that uses a local basis function is a nite-element
method. A series expansion method that uses a global basis function orthogonal to
the residual is a spectral method. A common nite-element method is the Galerkin
nite-element method. With this method, the local basis function is also orthogonal
to the residual. The basis functions of other nite-element methods may or may
not be orthogonal to the residual. Below, the Galerkin nite-element method is
discussed, and a type of spectral method is briey described.
6.5.1 Finite-element method
With the Galerkin nite-element method, basis functions are treated like weight
functions in that they weight a residual at each of several nodes along a spatial grid
(test space). The sum, over the test space, of the residual multiplied by the weight
is zero (e.g., Pepper et al. 1979; Celia and Gray 1992). Thus,
x
R
i
(x)e
i
(x)dx = 0 (6.72)
The purpose of this constraint is to minimize R
i
(x) by forcing its weighted average
over the domain to zero. With the Galerkin method, the weight function e
i
(x)
and test space in (6.72) are the same as the basis function e
j
(x) and trial space,
respectively, in (6.71). For other methods, e
i
(x) and e
j
(x) may differ, and the test
space may differ from the trial space. Methods in which the test and trial spaces
differ are PetrovGalerkin methods.
The residual of (6.70) is
R
i
(x) =
N
i
(x)
t
+u
N
i
(x)
x
N
i
t
+u
N
i
x
N
i
(x)
t
+u
N
i
(x)
x
0 (6.73)
193
which is the difference between the approximate and exact form of the equation.
Substituting (6.73) and (6.71) into (6.72) and assuming constant u gives
N
i
(x)
t
+u
N
i
(x)
x
e
i
(x)dx
=
j
N
j
e
j
(x)
+u

x
j
N
j
e
j
(x)
e
i
(x)dx
=
N
j
t
x
e
j
(x)e
i
(x)dx
+u
N
j
x
de
j
(x)
dx
e
i
(x)dx
= 0 (6.74)
Expanding the basis function in (6.74) over three nodes centered at i ( j = i
1, . . . , i +1), and taking a rst-order forward-difference approximation in time of
the result gives
N
i 1.t
N
i 1.th
h
x
i
x
i 1
e
i 1
(x)e
i
(x)dx +
N
i.t
N
i.th
h
x
i +1
x
i 1
e
i
(x)e
i
(x)dx
+
N
i +1.t
N
i +1.th
h
x
i +1
x
i
e
i +1
(x)e
i
(x)dx +u
N
i 1.t
x
i
x
i 1
de
i 1
(x)
dx
e
i
(x)dx
+ N
i.t
x
i +1
x
i 1
de
i
(x)
dxe
i
(x)dx + N
i +1.t
x
i +1
x
i
de
i +1
(x)
dx
e
i
(x)dx
= 0 (6.75)
One set of basis functions is the chapeau (hat) function,
e
i
(x) =
x x
i 1
x
i
x
i 1
x
i 1
x x
i
x
i +1
x
x
i +1
x
i
x
i
- x x
i +1
0 all other cases
(6.76)
These weightings favor the center cell (i) and decrease to zero at i 1 and i +1,
giving them the appearance of a peaked hat. The functions are the same for e
j
(x).
When chapeau functions are used, the integral in the rst term of (6.75) simplies
to
x
i
x
i 1
e
i 1
(x)e
i
(x)dx =
x
i
x
i 1
x
i
x
x
i
x
i 1
x x
i 1
x
i
x
i 1
dx =
x x
i1
6
(6.77)
After other terms have been integrated, (6.75) becomes
(N
i 1.t
N
i 1.th
)Lx
i
+(N
i.t
N
i.th
)2(Lx
i +1
+Lx
i
) +(N
i +1.t
N
i +1.th
)Lx
i +1
6h
+u
N
i +1.t
N
i 1.t
2
= 0 (6.78)
where Lx
i
= x
i
x
i 1
and Lx
i +1
= x
i +1
x
i
.
194
i1 i i+1
u
i1
u
i
u
i+1
x
i
x
i+1
N
i
, K
i
, r
a,i
Figure 6.6 Locations of vari-
ables along a westeast grid
for a Galerkin nite-element
scheme.
Equation (6.75) assumes that u is constant. When u varies, its trial function is
U
i
(x) =
j
u
j
e
j
(x). A similar equation applies to the eddy diffusion coefcient.
Toon et al. (1988) and Pepper et al. (1979) show a Galerkin method with cha-
peau functions as nite-elements and a CrankNicolson time-stepping scheme. The
method uses a fourth-order approximation in space and a second-order approxi-
mation in time and solves the unidirectional advectiondiffusion equation given in
(6.1). The method allows variable scalar velocities, grid spacing, and eddy diffusion
coefcients. The solution is found by setting up i = 2. . . . . I 1 equations of the
form,
(N
i 1.t
N
i 1.th
)Lx
i
+(N
i.t
N
i.th
)2(Lx
i +1
+Lx
i
) +(N
i +1.t
N
i +1.th
)Lx
i +1
h
+j
c
(
i
N
i +1
+
i
N
i

i
N
i 1
)
t
+(1 j
c
)(
i
N
i +1
+
i
N
i

i
N
i 1
)
th
= 0
(6.79)
where j
c
= 1, 0.5, or 0 implies an implicit, CrankNicolson, or forward Euler
solution, respectively, and
i
= (u
i
+2u
i 1
) +
3(
a.i
K
i
+
a.i 1
K
i 1
)
a.i 1
Lx
i
i
= (u
i 1
+u
i +1
) +
3[(
a.i 1
K
i 1
+
a.i
K
i
)Lx
i +1
+(
a.i
K
i
+
a.i +1
K
i +1
)Lx
i
]
a.i
Lx
i
Lx
i +1
i
= (u
i
+2u
i +1
)
3(
a.i
K
i
+
a.i +1
K
i +1
)
a.i +1
Lx
i +1
(6.80)
Velocities, eddy diffusion coefcients, and densities in (6.80) are located at grid-
cell boundaries, as shown in Fig. 6.6. Equation (6.79) can be written in tridiagonal
form as
A
i
N
i 1.t
+ B
i
N
i.t
+ D
i
N
i +1.t
= E
i
N
i 1.th
+ F
i
N
i.th
+ G
i
N
i +1.th
(6.81)
195
i = 1 i = 2 i = 3
u
1
u
2
u
3
x
2
x
3
N
1
, K
1
, r
a,1
(a)
i = I2 i = I1 i = I
u
I2
u
I1
u
I
x
I1
x
I
N
I
, K
I
,

r
a,I
(b)
Figure 6.7 Location of grid variables at (a) low and (b) high boundaries.
where
A
i
= Lx
i
hj
c
i
E
i
= Lx
i
+(1 j
c
)
i
B
i
= 2(Lx
i
+Lx
i +1
) +hj
c
i
F
i
= 2(Lx
i
+Lx
i +1
) h(1 j
c
)
i
D
i
= Lx
i +1
+hj
c
i
G
i
= Lx
i +1
h(1 j
c
)
i
(6.82)
Like (6.45), Equation (6.81) may be solved with a tridiagonal matrix technique.
Figure 6.7 shows locations of variables for boundaries on a limited-area grid.
The coefcients for outow from the high boundary are
A
I
= Lx
I
hj
c
I
E
I
= Lx
I
+h(1 j
c
)
I
B
I
= 2Lx
I
+hj
c
I
F
I
= 2Lx
I
h(1 j
c
)
I
D
I
= 0 G
I
= 0
(6.83)
where
I
= u
I
+2u
I1
+
3(
a.I
K
I
+
a.I1
K
I1
)
a.I1
Lx
I
I
= 4u
I
u
I1
+
9
a.I
K
I
+3
a.I1
K
I1
a.I
Lx
I
(6.84)
The coefcients for outow from the low boundary are
A
1
= 0 E
1
= 0
B
1
= 2Lx
1
+hj
c
1
F
1
= 2Lx
2
h(1 j
c
)
1
D
1
= Lx
2
+hj
c
1
G
1
= Lx
2
h(1 j
c
)
1
(6.85)
where
1
= 4u
1
+u
2
+
9
a.1
K
1
+3
a.2
K
2
a.1
Lx
2
1
= u
1
+2u
2
3(
a.1
K
1
+
a.2
K
2
)
a.2
Lx
2
(6.86)
Inow boundary equations are obtained by extending (6.81) one node beyond
the boundary and estimating concentrations and wind speeds in this virtual node.
Virtual-node values may be set to values just inside the boundary or extrapolated
from two or three nodes inside the boundary. Another option is to set virtual-node
196
200
0
200
400
600
800
1000
1200
0 5 10 15 20 25 30
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g
e
n
e
r
i
c
)
Grid-cell number
(a)
200
0
200
400
600
800
1000
1200
0 5 10 15 20 25 30
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g
e
n
e
r
i
c
)
Grid-cell number
(b)
400
200
0
200
400
600
800
1000
1200
0 5 10 15 20 25 30
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g
e
n
e
r
i
c
)
Grid-cell number
(c)
Figure 6.8 Preservation of a Gaussian peak during nite-element transport over
a grid with periodic boundary conditions when (a) uhLx = 0.02, (b) uhLx =
0.25, (c) uhLx = 0.6. Solid lines are initial values, short-dashed lines are values
after four revolutions, and long-dashed lines are values after eight revolutions.
values equal to an average of values just inside the boundary from the current and
the previous time step.
Equation (6.81) can also be solved with periodic boundary conditions. Figures
6.8 (a)(c) show results from the nite-element method described above when a
Gaussian plume was advected over a grid with such boundary conditions. Grid
spacing and wind speeds were uniform for the test. Each gure shows a result
when a different value of uhLx was used. The gures indicate that uhLx should
be 0.25 or less to minimize numerical diffusion, which is the articial spreading of
the peak, and numerical dispersion, which is oscillations upwind and downwind
of the peak. If grid spacing is 5 km and the average wind speed is 10 m s
1
, the
time step in this case should be no larger than 0.25 5000 m/10 m s
1
= 125 s.
6.5.2 Pseudospectral method
Like the nite-element method, the pseudospectral method involves the replace-
ment of the spatial differential operator by a nite series of basis functions. In the
case of the nite-element method, the basis functions are local functions. In the case
of the pseudospectral method, the basis functions are a nite series of orthogonal
functions. The difference between a spectral and a pseudospectral method is that,
with the former, time and space derivatives are approximated with a nite series.
197
With the latter, spatial derivatives are approximated with a nite series, but time
derivatives are approximated with an explicit Taylor series expansion or another
method.
If wind speed in the x-direction is constant, the westeast advection equation
for gas number concentration is
N
t
+u
N
x
= 0 (6.87)
The pseudospectral solution to this equation can be found by representing N(x,t)
over the interval 0 x L by the Fourier series
N(x. t) =
k=0
a
k
(t) e
ik2xL
(6.88)
(Orszag 1971; Wengle and Seinfeld 1978; Hack 1992), where k is the wavenumber
and a
k
(t) are complex Fourier coefcients. At t = 0, N is a known function of x.
Values of a
k
(0) are found by integrating both sides of (6.88) from 0 x L. The
result is
a
k
(0) =
1
L
L
0
N(x. 0)e
ik2xL
dx (6.89)
For practical application, the innite series in (6.88) is truncated to a nite number
of wavenumbers, K, giving
N(x. t) =
K
k=0
a
k
(t)e
ik2xL
(6.90)
The larger the value of K, the more accurate the estimate of N.
A pseudospectral solution to (6.87) can be found by taking a second-order,
central-difference approximation of (6.90) with respect to time and the partial
derivative of (6.90) with respect to space. The resulting expressions are
N
t

1
2h
k=0
a
k.t
e
ik2xL
k=0
a
k.t2h
e
ik2xL
(6.91)
N
x
=
K
k=0
ik2a
k.th
L
e
ik2xL
(6.92)
respectively. Substituting these into (6.87) yields
1
2h
K
k=0
(a
k.t
a
k.t2h
)e
ik2xL
= u
K
k=0
ik2a
k.th
L
e
ik2xL
(6.93)
which can be separated into K equations of the form
a
k.t
a
k.t2h
2h
=
uik2a
k.th
L
(6.94)
Equation (6.94) is explicit and can be solved immediately, since values of a
k
at time
t h and t 2h are known from previous time steps. Fourier coefcients for the
198
6.7 Advection schemes used in air-quality models
rst time step (t = h) are found by taking a forward- instead of central-difference
approximation in (6.91). Once Fourier coefcients have been determined from
(6.94), they are substituted back into (6.90) to give an estimate of N at time t for
any value of x.
An advantage of a pseudospectral scheme over a nite-difference approximation
is that only K equations need to be solved in the pseudospectral scheme. In a nite-
difference scheme, I nite-difference equations need to be solved per time step,
where I is the number of grid cells in one direction. Usually, I > K. Whereas the
pseudospectral solution to the linear advection problem is easy to implement, it is
not readily applied to nonlinear problems, such as when u varies in space or when
u is a prognostic variable. In such cases, a separate basis function for u is required.
The multiplication of two nite series, such as one for u and one for N, results in
additional terms, slowing the pseudospectral numerical solution. One way to avoid
the multiplication of spectral-basis-function products is with the spectral transform
method (Eliasen et al. 1970; Orszag 1970).
For global modeling, the basis functions used are the spherical harmonics. These
functions are a combination of sine and cosine functions along the zonal (west
east) direction and Legendre functions along the meridianal (southnorth) direction
on a sphere. They are computationally fast in comparison with some other basis
functions. Spectral and pseudospectral methods are discussed in more detail in
Orszag (1970), Washington and Parkinson (1986), Holton (1992), Hack (1992),
and Krishnamurti et al. (1998). Pseudospectral techniques are commonly used to
discretize horizontal advection terms in global models.
6.6 FINITE-VOLUME METHODS
Finite-volume methods are methods of solving partial differential equations that
divide space into discrete volumes. Partial differential equations are integrated
over each volume. Variables are then approximated in each volume by averaging
across the volume. Variables are exchanged between neighboring volumes at vol-
ume interfaces in a ux-conserving manner. Finite-volume methods are generally
applied over irregular grids (Section 4.1). A detailed discussion of the nite-volume
method is given in Durran (1999).
6.7 ADVECTION SCHEMES USED IN AIR-QUALITY MODELS
Advection schemes are useful in three-dimensional air quality models only if they
are not allowed to generate unphysical extreme concentrations (they are bounded),
do not generate articial oscillations (nonoscillatory), and preserve gradients of
mixing ratio (monotonic). When a scheme is bounded, the mixing ratio in the
current grid cell can never increase above or decrease below the mixing ratio in
any adjacent grid cell after an advection time step. When such a condition is satised
over a grid domain as a whole, no mixing ratio can fall below the lowest or rise
above the highest initial mixing ratio anywhere in the domain. Schemes that are
bounded are generally nonoscillatory and monotonic as well.
199
A problem with nonbounded schemes is that they can produce negative mix-
ing ratios (or concentrations), which are not physical, or mixing ratios above the
maximum possible, which may lead to inaccurate predictions of pollution lev-
els. Figure 6.8, for example, shows a scheme in which concentrations drop below
zero. This scheme, in its original form, is unbounded and oscillatory. The scheme
can be made bounded by limiting the minimum and maximum mixing ratios
upwind and downwind of each grid cell, then adjusting uxes in the horizontal
to conserve mass. Although such adjustments may increase numerical diffusion
for some schemes, they remove oscillations, spurious peaks, and negative mixing
ratios.
In previous sections, several schemes used for solving unidirectional advection
diffusion equations were given. Below, additional schemes are described. Some of
the schemes, as originally implemented, are bounded, nonoscillatory, and mono-
tonic, whereas others are not.
The forward Euler TaylorGalerkin method (Donea 1984) is a Galerkin method
that uses the forward Euler rather than the CrankNicolson time discretization. A
Galerkin method that uses a chapeau basis function but a modied weighting func-
tion is the PetrovGalerkin method (Hughes and Brooks 1979). The accurate space
derivative (ASD) scheme of Gazdag (1973) is a pseudospectral scheme in which time
and spatial derivatives are approximated with truncated Taylor series expansions
and truncated Fourier series expansions, respectively. The scheme of Smolarkiewicz
(1983) is an iterative, positive-denite, backward-difference approximation scheme
that reduces numerical diffusion by correcting the velocity with an antidiffusion
velocity, derived from the truncated terms in a Taylor-series expansion of the
advection equation.
Collela and Woodward (1984), Prather (1986), Tremback et al. (1987), Bott
(1989), Carpenter et al. (1990), and Easter (1993) derived solutions to the
advectiondiffusion equation in which mixing ratios are expanded in space with
parabolic functions or higher-order polynomials. Each polynomial is derived from
mixing ratios in several adjacent grid cells. In the case of Prather (1986), rst and
second derivatives of the parabolic function are stored for use during subsequent
advection time steps. These schemes generally modify uxes at grid-cell interfaces
to ensure mass conservation. Related schemes include those by Yamartino (1993),
Rasch (1994), and Thuburn (1996, 1997).
Walcek and Aleksic (1998) and Walcek (2000) developed a bounded, nonoscil-
latory, and monotonic scheme in which spatial gradients in mixing ratio are tted
with linear functions, rather than polynomials, optimized to preserve local peaks
and reduce numerical diffusion. Fluxes across grid-cell interfaces are limited to
ensure mass conservation. Figure 6.9 shows an example of the conservation and
low numerical diffusion resulting from this scheme when a tracer with mixing
ratio represented by several shapes is advected on a two-dimensional grid. The
smooth background eld is a product of the bounded and nonoscillatory nature
of the scheme. The relative preservation of the shapes is an indicator of the highly
monotonic nature of the scheme.
200
6.7 Advection schemes used in air-quality models
1
2
0
1
0
0
8
0
6
0
4
0
2
0
0
Z
1
0
0
8
0
6
0
4
0
Y
2
0
2
0
4
0
X
6
0
8
0
1
0
0
100
100
100
100
Max= 100
(a)
1
2
0
1
0
0
8
0
6
0
4
0
2
0
0
Z
1
0
0
8
0
6
0
4
0
Y
2
0
2
0
4
0
X
6
0
8
0
1
0
0
99
54
100
13
Max= 100
(b)
Figure 6.9 Comparison of (a) initial and expected exact nal shapes and (b)
actual nal shapes after six rotations with 628 time steps per rotation of advec-
tion around a two-dimensional 100 100 horizontal domain center using peri-
odic boundary conditions. The maximum Courant number was near 0.5. The
numbers represent peak mixing ratios for each shape. The scheme and diagram
originate from Walcek (2000).
In the trajectory grid scheme (Chock et al. 1996), advection is solved with a fully
Lagrangian method and eddy diffusion is solved with an Eulerian diffusion scheme.
In the scheme of Nguyen and Dabdub (2001), spatial derivatives of a tracer are
approximated with quintic splines and time derivatives are approximated with a
Taylor series expansion.
201
Many of the schemes described above have been compared with each other
when used alone (e.g., Chock 1991; Walcek 2000) or coupled with chemistry (e.g.,
Chock and Winkler 1994; Dabdub and Seinfeld 1994).
6.8 SUMMARY
In this chapter, methods of solving partial differential equations, and specically
the advectiondiffusion equation, were discussed. The methods include nite-
difference, series expansion, and nite-volume methods. Anite-difference approx-
imation involves the replacement of a continuous differential operator with a
discrete difference analog along a predetermined number of spatial or temporal
nodes. The order of approximation increases with the number of nodes along
which the differential is discretized. Forward-, backward-, and central-difference
approximations of the rst and second derivatives of a variable were derived for a
variety of orders under the assumption of constant grid spacing. Finite-difference
approximations of the advectiondiffusion equation were also discussed for cases
of nonuniform grid spacing and variable eddy diffusion coefcients. The semi-
Lagrangian method, which is in the family of nite-difference methods, was also
discussed, and series expansion methods were described. These methods, which
include nite-element and pseudospectral methods, involve the replacement of a
dependent variable with a nite series that approximates the variable. In addition,
several time-stepping schemes and their stability characteristics were described.
Finally, characteristics necessary for schemes solving the advectiondiffusion equa-
tion in three-dimensional air pollution models were discussed.
6.9 PROBLEMS
6.1 Identify ve characteristics of a good numerical approximation, and explain
which of these characteristics you think the forward Euler scheme (Section
6.4.4.2) with no diffusion has. (Hint: the characteristics of the scheme may be
obtained by estimation or by writing a one-dimensional code with periodic
boundary conditions and testing the effect of changes in time step and grid
spacing).
6.2 If an implicit approximation such as (6.36) is unconditionally stable, why
is it less accurate, for the same grid spacing and time step, than (6.44) when
j
c
= 0.5?
6.3 What advantage does the nite-element scheme of fourth-order in space and
second-order in time shown in this chapter have over a nite-difference
scheme of the same orders?
6.4 Find a sixth-order central-difference approximation of Nx. Assume
constant grid spacing.
6.5 Find a fth-order backward-difference approximation of Nx . Assume
constant grid spacing.
6.6 Suppose the horizontal wind in a large domain shifted from a 5 ms
1
south-
westerly wind to a 10 ms
1
westerly wind to an 8 ms
1
northwesterly wind
202
back to a 10 ms
1
westerly wind wind 3, 2, and 1 hour ago, respectively.
Draw the trajectory of a Lagrangian pollution parcel carried by the wind,
and calculate the straight-line distance the parcel traveled between 4 hours
ago and the present time. Ignore curvature of the Earth.
6.7 Set up a westeast grid with periodic boundary conditions that has 100
uniformly spaced grid cells. Assume the last grid cell to the east meets the
rst grid cell to the west. Set Lx = 5 km in each cell and u = 5 m s
1
at
each cell westeast boundary. Initialize the gas concentration on the grid
with the Gaussian distribution N
i
= N
0
e
(i50)
2
8
, where N
0
= 10
12
molec.
cm
3
is a peak concentration and i corresponds to a cell boundary. Solve
(6.36), assuming the eddy diffusion coefcient is zero, with the matrix given
in (6.42). Use a time step of 5 s, and solve until 10 revolutions around the
grid have been completed. Plot concentration versus grid cell at the end of
each revolution. Discuss the numerical diffusion of the peak over time.
6.8 Do the same as in Problem 6.7, but instead solving with the fourth-order
RungeKutta technique. Use a time step of 0.1 s to obtain an exact solution.
Plot the difference between the exact and modeled concentration at one
point along the grid at a specic time during the simulation versus time step
when different time steps are used to demonstrate the fourth-order nature
of the solution.
203
7
Finite-differencing the equations
of atmospheric dynamics
M
any numerical models have been developed to predict the weather on
regional and global scales. The rst was that of Richardson (1922), which
was solved by hand (Section 1.1). The rst computer model was the one-
dimensional model of Charney (1949), developed as part of the project to apply
the worlds rst computer (Section 1.1). Three-dimensional computer models today
usually discretize time and vertical spatial derivatives with nite-difference approx-
imations. Horizontal advection terms are discretized with nite-difference, spectral,
nite-element, or semi-Lagrangian approximations. In this chapter, one numeri-
cal solution to the equations of atmospheric dynamics, a three-dimensional nite-
difference solution in spherical-sigma-pressure coordinates, is discussed. The equa-
tions solved include the continuity equation for air, species continuity equation,
thermodynamic energy equation, horizontal momentumequations, and hydrostatic
equation. The solution scheme originates from Arakawa and Lamb (1977) and
Arakawa and Suarez (1983). Although more recent and advanced versions of the
scheme have been developed, and although many other numerical solutions exist,
the version presented was chosen for illustration purposes. A modeling project at
the end of the chapter allows students to develop a basic regional- or global-scale
dynamics model from the equations given.
7.1 VERTICAL MODEL GRID
A complete weather-prediction model requires all the components shown in
Fig. 1.1. In this chapter, only the meteorological component is addressed. Anumeri-
cal technique is described to predict wind speed and direction, air pressure, energy,
and moisture. Techniques for solving equations for other processes are given in
other chapters of this text. Chapter 21 describes how different components of a
model are integrated and tested.
The rst step in the development of a dynamics module is to dene the vertical
coordinate system and location of variables within this system. For the present
application, the sigma-pressure vertical coordinate will be used. Figure 7.1 shows
the location of model variables in the vertical with this coordinate system. Vertical
scalar velocities are located at the top and bottom of each layer. Horizontal scalar
velocities and other variables (e.g., gas and particle mixing ratios and potential
virtual temperature) are located at the vertical midpoint of the layer. This type
204
7.1 Vertical model grid
_______Model top boundary______
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
________________________________
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
________________________________
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
________________________________
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
________________________________
_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
______Model bottom boundary____
s
1/2
= 0, s
1/2
= 0, p
a,top
q
1
, q
v,1
, u
1
, v
1
, p
a,1
s
1+1/2
, s
1+1/2
, p
a,1+1/2
q
2
, q
v,2
, u
2
, v
2
, p
a,2
s
k1/2
, s
k1/2
, p
a ,k1/2
q
k
, q
v,k
, u
k
, v
k
, p
a,k
s
k+1/2
, s
k+1/2
, p
a,k+1/2
q
k+1
, q
v,k+1
, u
k+1
, v
k+1
, p
a,k+1
s
N
L
1/2
, s
N
L
1/2
, p
a, N
L
1/2
q
N
L
, q
v,N
L
, u
N
L
, v
N
L
, p
a, N
L
s
N
L
+1/2
= 0, s
N
L
+1/2
= 1, p
a,surf
Figure 7.1 Location of variables in the vertical direction of a
sigma-pressure coordinate Lorenz grid. Integer subscripts cor-
respond to layer centers, and fraction subscripts correspond to
layer boundaries.
of grid is called the Lorenz grid, after Lorenz (1960). A similar grid that stores
potential virtual temperature at layer top and bottomboundaries instead of centers
is called the CharneyPhillips grid, after Charney and Phillips (1953). Both grids
are described in more detail in Arakawa and Konor (1995). In Fig. 7.1, layer center
indices increase from 1, at the top of the model, to N
L
, at the bottom, and layer
boundary indices increase from 12, at the top, to N
L
+ 12 at the bottom.
Sigma-pressure coordinate vertical scalar velocities at the bottom of the model
are set to zero, since the model bottom is assumed to be a material surface, namely,
the ground surface. Setting the model bottom vertical scalar velocities to zero also
helps to lter out Lamb waves, which are horizontally propagating acoustic waves
that give rise to pressure, velocity, and density perturbations (Chapter 4). Since the
amplitude of Lamb-wave oscillations is usually greatest at the lowest model level,
setting the vertical scalar velocity to zero at that level decreases perturbations due
to the oscillations.
At the model top, vertical scalar velocities are also set to zero for lack of a better
assumption. Zero velocities at the top are not realistic, unless the model top is the
top of the atmosphere. An alternative to setting top vertical scalar velocities to zero
is to estimate them, which is also an error-prone process. In either case, errors can
be reduced by raising the model top to a higher altitude. This may require adding
more model layers, increasing computational requirements.
In the sigma-pressure coordinate, the -thickness of a layer is the difference in
-values between the top and bottom of the layer. Thus,
L
k
=
k+12
k12
for k = 1. . . . . N
L
(7.1)
where the -value at the top and bottomof each layer is predetermined and constant
throughout the model domain and during the simulation. The -value at the top
of the model usually equals zero (
12
=
top
= 0), and that at the ground surface
205
Finite-differencing the equations of atmospheric dynamics
usually equals unity (
N
L
+12
=
surf
= 1) although these limits are arbitrary. The
setting of -values at the boundaries of other layers is described below.
When a model extends fromthe ground to the top of the stratosphere, the choice
of -values is important for maintaining numerical stability. In the stratosphere,
small pressure changes correspond to large altitude changes. Near the surface, small
pressure changes correspond to small altitude changes. Thus, the - (and pressure-)
thickness of a layer near the model top should not be too large and that near the
ground should not be too small.
One way to select -values is rst to set up a test column and assume that
the altitude thickness of all layers in the test column is the same. To set up the
test column, it is necessary to dene the model top pressure, p
a,top
(hPa), which
is uniformly constant throughout the model domain and over time, and a mean
surface altitude z
surf,test
(m), which is just an average topographical altitude above
sea level over the model domain. A typical top pressure for a regional model is
250 hPa. That for a global model is 0.5 hPa. As a default, the mean surface altitude
can be set to zero.
From the model top pressure, the model top altitude (m) in the test column can
be estimated with
z
top,test
= z
below
+
p
a,below
p
a,top
a,below
g
below
(7.2)
derived from (2.41), where p
a,below
(Pa),
a,below
(kg m
3
) and g
below
(m s
2
) are
air pressure, air density, and gravity corresponding to the air pressure in Appendix
Table B.1 just greater than (at a lower altitude than) p
a,top
(Pa). The altitude at the
bottom of each test column layer can then be calculated from
z
k+12.test
= z
surf,test
+(z
top,test
z
surf,test
)
1
k
N
L
for k = 0. . . . . N
L
(7.3)
where k is the layer index number. The pressure p
a.k+12.test
at each altitude z
k+12.test
is then determined from (2.41) using interpolated density and gravity values from
Appendix Table B.1. Finally, the -value at the bottom boundary of each model
layer is calculated with
k+12
=
p
a.k+12.test
p
a,top
p
a.N
L
+12.test
p
a,top
for k = 1. . . . . N
L
(7.4)
At the model top (k = 0), p
a.12.test
= p
a,top
. The -value at the boundary of each
real model layer is then set equal to that at each test layer.
Once -values have been determined, they are used to calculate the pressure
at the bottom boundary of each model layer. This can be done only after surface
pressure p
a,surf
(Pa) has been dened for each column in the model. For realistic sim-
ulations, surface pressures should be initialized with interpolated measurements.
For idealized simulations in the absence of topography, surface pressure can be
initialized everywhere with standard sea-level surface pressure (1013.25 hPa). In
idealized simulations with topography, topographical surface air pressure can be
206
7.1 Vertical model grid
estimated by applying (2.41) between sea level (z = 0) and the topographical alti-
tude using air density and gravity estimates from Appendix Table B.1. From the
model surface and top pressures, the pressure thickness of each model column is
dened as
a
= p
a,surf
p
a,top
(7.5)
The pressure at the bottom boundary of each model layer is then
p
a.k+12
= p
a,top
+
k+12
a
(7.6)
Although p
a,top
is the same in each model column, p
a,surf
differs for each column;
thus,
a
and p
a.k+12
also differ for each column and grid cell, respectively.
The vertical midpoint of a layer can be dened as the center of height, the center
of mass, or the pressure at which the mass-weighted mean of a variable is located.
If it is dened as the center of mass, the air pressure at the vertical center is
p
a.k
= p
a.k12
+0.5( p
a.k+12
p
a.k12
) (7.7)
Alternatively, the vertical midpoint can be dened such that the nite-difference
expression for the hydrostatic equation, given in Section 7.6, becomes exact for an
isentropic atmosphere (e.g., where potential virtual temperature increases mono-
tonically with height, as shown in Fig. 2.13). In such a case, the midpoint is the
pressure at which the mass-weighted mean of P, the potential temperature factor
dened in (2.98), is located (Arakawa and Suarez 1983). The mass-weighed mean
of P in layer k is
P
k
=
1
p
a.k+12
p
a.k12
p
a.k+12
p
a.k12
P dp
a
=
1
1 +
P
k+12
p
a.k+12
P
k12
p
a.k12
p
a.k+12
p
a.k12
(7.8)
where values of p
a
and P at layer boundaries are p
a.k+12
and
P
k+12
=
p
a.k+12
1000 hPa
(7.9)
respectively. Once P
k
has been found from (7.8), the pressure corresponding to
P
k
is
p
a.k
=
1000 hPa
P
1
k
(7.10)
207
In the grid described in Fig. 7.1, potential virtual temperature is evaluated at the
vertical midpoint of a layer. However, values are needed at layer tops and bottoms
as well to calculate vertical energy uxes. If (7.10) is used to locate the vertical
pressure-midpoint of a layer, a consistent formulation for
v
at a layer boundary
is
v.k+12
=
(P
k+12
P
k
)
v.k
+(P
k+1
P
k+12
)
v.k+1
P
k+1
P
k
(7.11)
Example 7.1
Given the following pressure and potential virtual temperature proles, cal-
culate the pressure at the midpoints, with respect to mass and to the mass
integral of P, of layers k and k+1 and potential virtual temperature at the
layer boundary, k +12. Assume the air is dry.
p
a.k12
= 700 hPa
v.k
= 308 K
p
a.k+12
= 750 hPa
v.k+1
= 303 K
p
a.k+32
= 800 hPa
SOLUTION
Substituting values above into (7.7) gives p
a.k
= 725 hPa and p
a.k+1
= 775 hPa.
From (7.9), P
k12
= 0.9030214, P
k+12
= 0.9210167, and P
k+32
= 0.9381747.
Substituting these into (7.8) gives P
k
=0.9120929 and P
k+1
=0.9296616. From
(7.10), p
a.k
= 724.897 hPa and p
a.k+1
= 774.904 hPa. In sum, pressure at the
midpoint with respect to the mass integral of P is slightly lower than pressure
at the mass center of a layer. From (7.11),
v.k+12
= 305.54 K.
7.2 THE CONTINUITY EQUATION FOR AIR
In the sigma-pressure coordinate, the continuity equation for air is used prognosti-
cally to calculate changes in total column pressure and diagnostically to calculate
vertical scalar velocities at the top and bottom of each grid cell. From (5.66), the
prognostic equation for the change in column pressure was
R
2
e
cos
a
t
1
0
e
(u
a
R
e
) +

(v
a
R
e
cos )
d (7.12)
Replacing d, dt, d
e
, d, and d with L, h, L
e
, L, and L, respectively, and
multiplying through by L
e
L give a rst-order in time and second-order in space
208
7.2 The continuity equation for air
j+1
j+1/2
j
j1/2
j1
j3/2
i3/2
i1
i1/2
i
i+1/2
i+1
a

a
a
v
u
u
u
u
u
u
u
u
u
v
v
v
v
v
v
v
v
G
G
F F
j1
j1/2
j3/2
j
j+1/2
j+1
u
u
u
v
v
v
i3/2
j+3/2
j+3/2
i+1
i+3/2
i+3/2
i i1/2
i1
i+1/2
Figure 7.2 The Arakawa C-grid. The grid is centered around
a
for
nite-differencing of the continuity equation. The gure shows exact
locations of u, v, and
a
relative to lateral boundaries.
approximation for the column pressure as
R
2
e
cos L
e
L
i. j
a.t
a.th
h
i. j
=
N
L
k=1
(u
a
R
e
LL
e
L)
i +12. j
(u
a
R
e
LL
e
L)
i 12. j
L
e
k.th
N
L
k=1
(v
a
R
e
cos LL
e
L)
i. j +12
(v
a
R
e
cos LL
e
L)
i. j 12
L
k.th
(7.13)
The horizontal nite-difference grid chosen for discretizing this equation is the
Arakawa C-grid (Arakawa and Lamb 1977), shown in Fig. 7.2. In this grid, column
pressures (
a
) are evaluated at
a
-points, which are bounded to the west and east
by u-scalar velocity points and to the south and north by v-scalar velocity points.
The left sides of the second and third numerators in (7.13) are uxes entering the
solid box shown in Fig. 7.2, and the right sides are uxes leaving the box.
The grid shown in Fig. 7.2 is a limited-area grid, which is a grid with lateral
boundaries; u-points lie on western and eastern boundaries, and v-points lie on
southern and northern boundaries. Boundary conditions for a limited-area grid
are discussed in this chapter. If the limited-area grid is nested within a global grid,
boundary conditions for the limited-area grid are obtained from the global grid,
as described in Section 21.1.11. A global grid converges to a singularity at the
poles in the southnorth direction. As such, the poles must be treated carefully;
otherwise mass converges at the poles and computational instabilities arise.
209
One method of treating the poles on a global grid is to set v-scalar velocities to
zero at the southernmost and northernmost boundaries. As a result, winds advect
from west to east or east to west around the poles (u = 0), avoiding the singular-
ity. In the westeast direction on a global grid, boundary conditions are periodic
(Section 6.4.4.2), so values on the west boundary are the same as values on the
east boundary at u-, v-, and
a
-points.
Equation (7.13) can be solved to give a prognostic equation for the column
pressure,
a.i. j.t
=
a.i. j.th
R
2
e
cos L
e
L
i. j
N
L
k=1
[(F
i +12. j
F
i 12. j
+ G
i. j +12
G
i. j 12
)
k.th
L
k
] (7.14)
where F and G are uxes dened at u- and v-points, respectively, as shown in
Fig. 7.2. At interior points in the domain, F and G are
F
i +12. j.k.th
=
a.i. j
+
a.i +1. j
2
(uR
e
L)
i +12. j.k
th
(7.15)
G
i. j +12.k.th
=
a.i. j
+
a.i. j +1
2
(vR
e
cos L
e
)
i. j +12.k
th
(7.16)
where the division by two indicates that the
a
values at i, j points have been
averaged to obtain approximate
a
values at i + 12, j and i, j + 12 points. At
eastern and northern lateral boundaries, F and G are
F
I+12. j.k.th
= [
a.I. j
(uR
e
L)
I+12. j.k
]
th
(7.17)
G
i.J +12.k.th
= [
a.i.J
(vR
e
cos L
e
)
i.J +12.k
]
th
(7.18)
respectively, where I + 12 and J + 12 are the easternmost and northernmost u-
and v-points, respectively (i.e., points i +32 and j +32, respectively, in Fig. 7.2).
Similar equations can be written for the western and southern boundaries. Once
new column pressures have been calculated from (7.14) at the end of a time step,
they are used in (7.6) to obtain new pressures at the top and bottom boundaries
of each grid cell. Layer vertical midpoint pressures are then obtained from (7.8)
(7.10).
After new column pressures have been determined, vertical scalar velocities at
the top and bottomof each grid cell are calculated diagnostically from(5.69), given
as

a
R
2
e
cos =
e
(u
a
R
e
) +

(v
a
R
e
cos )
d
R
2
e
cos
a
t
(7.19)
Replacing differential and integral operators in this equation with nite-difference
analogs and multiplying through by L
e
L yield the vertical scalar velocity at the
210
7.3 The species continuity equation
bottom of a layer as

a
R
2
e
cos L
e
L
i. j.k+12.t
=
k
l=1
(u
a
R
e
L
e
LL)
i 12. j
(u
a
R
e
L
e
LL)
i +12. j
L
e
l.th
l=1
(v
a
R
e
cos L
e
LL)
i. j 12
(v
a
R
e
cos L
e
LL)
i. j +12
L
l.th
k+12
R
2
e
cos L
e
L
i. j
a.t
a.th
h
i. j
(7.20)
Substituting uxes from (7.15)(7.18) into (7.20) gives the vertical scalar velocity
as

i. j.k+12.t
=
1
a
R
2
e
cos L
e
L
i. j.t
l=1
[(F
i +12. j
F
i 12. j
+ G
i. j +12
G
i. j 12
)
l.th
L
l
]
k+12
a.t
a.th
h
a.t
i. j
(7.21)
Equation (7.21) is solved in the order, k =1, . . . , N
L
1. At k =0,
12
=
top
= 0
by denition, and at k = N
L
, the equation predicts a vertical scalar velocity at the
lowest model boundary of zero (
N
L
+12
= 0). This characteristic is demonstrated
mathematically by combining (7.14) and (7.21) with k =N
L
in the latter equation.
7.3 THE SPECIES CONTINUITY EQUATION
Transport of each gas and aerosol-particle component in a model is simulated
with the species continuity equation. This equation can be solved as three one-
dimensional equations, a two-dimensional and a one-dimensional equation, or
one three-dimensional equation. In Chapter 6, techniques for solving the one- and
two-dimensional equations were given. Here, an explicit solution to the three-
dimensional equation is shown.
The ux form of the species continuity equation in spherical-sigma-pressure
coordinates was given in (5.73) as
R
2
e
cos

t
(
a
q)
e
(u
a
qR
e
) +

(v
a
qR
e
cos )
+
a
R
2
e
cos
( q)
=
a
R
2
e
cos
(
a
K
h
)q
a
+
N
e.t
n=1
R
n
(7.22)
211
where q is the moist-air mass mixing ratio of the species (or specic humidity,
in the case of water vapor). Replacing differential operators with nite-difference
analogs and multiplying through by L
e
L give
R
2
e
cos L
e
L
i. j
a.t
q
t
a.th
q
th
h
i. j.k
+
(u
a
qR
e
L
e
L)
i +12. j.k.th
(u
a
qR
e
L
e
L)
i 12. j.k.th
L
e
+
(v
a
qR
e
cos L
e
L)
i. j +12.k.th
(v
a
qR
e
cos L
e
L)
i. j 12.k.th
L
+
a.t
R
2
e
cos L
e
L
(
t
q
th
)
k+12
(
t
q
th
)
k12
L
k
i. j
=
a
R
2
e
cos L
e
L
(
z
a
K
h
z
) q
a
+
N
e.t
n=1
R
n
i. j.k.th
(7.23)
where the eddy diffusion termis not differenced here for simplicity. In this equation,
a
at time t, calculated from (7.21), is required. Substituting (7.15)(7.18) into
(7.23) and interpolating qs to u- and v-points where necessary give a prognostic
form of the species continuity equation as
q
i. j.k.t
=
(
a
q)
i. j.k.th
a.i. j.t
+
h
a.t
R
2
e
cos L
e
L
i. j
F
i 12. j
q
i 1. j
+q
i. j
2
F
i +12. j
q
i. j
+q
i +1. j
2
+G
i. j 12
q
i. j 1
+q
i. j
2
G
i. j +12
q
i. j
+q
i. j +1
2
k.th
+
a.t
R
2
e
cos L
e
L
(
t
q
th
)
k12
(
t
q
th
)
k+12
L
k
i. j
+
a
R
2
e
cos L
e
L
(
z
a
K
h
z
) q
a
+
N
e.t
n=1
R
n
i. j.k.th
(7.24)
where the uxes are the same as those shown in Fig. 7.2. In this equation, q values
outside lateral boundaries (e.g., at (0, j), (I + 1, j) (i, 0), and (i, J + 1) points)
on a nonnested, limited-area grid can be specied, set to nearest q values inside
the boundary from the current time step, or set to nearest q values inside the
boundary from the previous time step. Arakawa (1984) and Lu (1994) discuss a
more detailed treatment of lateral boundary conditions in which uxes of scalars
at the boundaries are derived from a mass-conservation relationship.
212
7.4 The thermodynamic energy equation
Mixing ratios for (7.24) at the top and bottom of a layer can be interpolated
from layer midpoint values with
q
i. j.k12
=
ln q
i. j.k1
ln q
i. j.k
(1q
i. j.k
) (1q
i. j.k1
)
q
i. j.k+12
=
ln q
i. j.k
ln q
i. j.k+1
(1q
i. j.k+1
) (1q
i. j.k
)
(7.25)
respectively. Mixing ratios at the surface and model top are not needed, since = 0
at those locations.
7.4 THE THERMODYNAMIC ENERGY EQUATION
The nite-difference form of the thermodynamic energy equation is similar to that
of the species continuity equation. From(5.76), the ux formof the thermodynamic
energy equation in spherical-sigma-pressure coordinates was
R
2
e
cos

t
(
a
v
)
e
(u
a
v
R
e
) +

(v
a
v
R
e
cos )
+
a
R
2
e
cos
(
v
)
=
a
R
2
e
cos
a
K
h
)
v
a
+

v
c
p.d
T
v
N
e.h
n=1
dQ
n
dt
(7.26)
Replacing differential operators with nite-difference analogs, multiplying through
by L
e
L, taking the nite-difference of each termin (7.26), and substituting uxes
from (7.15)(7.18) give potential virtual temperature as
v.i. j.k.t
=
(
a
v
)
i. j.k.th
a.i. j.t
+
h
a.t
R
2
e
cos L
e
L
i. j
F
i 12. j
v.i 1. j
+
v.i. j
2
F
i +12. j
v.i. j
+
v.i +1. j
2
+G
i. j 12
v.i. j 1
+
v.i. j
2
G
i. j +12
v.i. j
+
v.i. j +1
2
k.th
+
a.t
R
2
e
cos L
e
L
(
t
v.th
)
k12
(
t
v.th
)
k+12
L
k
i. j
+
a
R
2
e
cos L
e
L
(
z
a
K
h
z
)
v
a
+

v
c
p.d
T
v
N
e.h
n=1
dQ
n
dt
i. j.k.th
(7.27)
where the eddy diffusion term is not differenced for simplicity.
v
values outside
lateral boundaries (e.g., at (0, j), (I + 1, j) (i, 0), and (i, J + 1) points) can be
specied, set to nearest
v
values inside the boundary from the current time step,
or set to nearest
v
values inside the boundary from the previous time step. The
213
parameters
v.i. j.k12
and
v.i. j.k+12
are potential virtual temperatures at the top
and bottom of a layer, respectively, found from (7.11). Values of
v
at the surface
and top are not needed, since = 0 at those locations.
7.5 THE HORIZONTAL MOMENTUM EQUATIONS
The horizontal momentum equations are nite-differenced with a scheme that is
more accurate than the differencing schemes described in the previous sections.
Following Arakawa and Lamb (1977), a second-order scheme is discussed that
conserves the domain-integrated kinetic energy from inertial processes and the
vertical components of absolute vorticity and enstrophy during advection by the
nondivergent part of horizontal velocity. Kinetic energy (J = kg m
2
s
2
) is
KE =
1
2
a
V(u
2
+v
2
) (7.28)
where V is a control volume (e.g., the volume of a grid cell). Absolute vorticity
(Section 4.3.5.4) is the sum of relative vorticity (
r
) and the Earths vorticity (f ).
The vertical component of absolute vorticity (s
1
) is
a.z
=
r.z
+ f =
v
x

u
y
+ f (7.29)
Enstrophy (s
2
) is one-half the square of absolute vorticity,
ENST =
1
2
2
a.z
(7.30)
Conserving kinetic energy, absolute vorticity, and enstrophy over a model domain
is important for reducing nonlinear computational instabilities. The kinetic energy
constraint reduces error by preventing computational cascades of kinetic energy
into small-scale motions (Arakawa and Lamb 1977). Conserving these properties
requires strict adherence to the horizontal interpolations presented below and per-
mits solutions to remain stable for long integration periods.
Figure 7.3 shows the conservation properties of the scheme of Arakawa and
Lamb (1977) described in this chapter, but written for two-dimensional nondi-
vergent (
h
v = 0), irrotational (f = 0) ow and in terms of the vorticity and
streamfunction equations in place of the horizontal momentum equations and the
continuity equation for air. The vorticity equation is
r.z
t +v
h

r.z
= 0, and
the streamfunction equation is
r.z
=
2
h
, where is streamfunction (m
2
s
1
),
which is related to u and v by u = y and v = x, respectively. Density
and temperature were assumed constant. The gure shows that the scheme con-
served vorticity, enstrophy, and kinetic energy in this case to about 10 decimal
places.
For three-dimensional divergent, rotational ow in spherical-sigma-pressure
coordinates, the ux form of the momentum equation in the u-direction from
214
7.5 The horizontal momentum equations
2 10
9
1 10
9
0 10
0
1 10
9
2 10
9
0 1200 2400 3600
R
e
l
a
t
i
v
e

e
r
r
o
r
Time from start (s)
Figure 7.3 Time-dependent change in relative
vorticity ( ) ---- (since f = 0), enstrophy (ENST)
-, and kinetic energy (KE) , summed over
the model domain, divided by the initial summed
values of these respective variables, for two-
dimensional nondivergent, irrotational ow in
the Cartesian coordinate, assuming constant
density and temperature. The equations solved
were the vorticity and streamfunction equations
(replacing the horizontal momentum equations
and the continuity equation for air) (Arakawa
and Lamb 1977). The fourth-order RungeKutta
scheme was used for time stepping. The grid
domain consisted of 16 16 grid cells, each
1.25 km 1.25 km on edge. The time step
was 10 s. The boundary conditions were periodic.
From Ketean and Jacobson (2005a).
(5.86) was
R
2
e
cos

t
(
a
u)
e
(
a
u
2
R
e
) +

(
a
uvR
e
cos )
+
a
R
2
e
cos
( u)
=
a
uvR
e
sin +
a
f vR
2
e
cos
R
e
a
+
e
+c
p.d
v
P
+ R
2
e
cos
a
(
a
K
m
) u (7.31)
The nite-difference form of this equation is obtained by assuming uxes enter or
leave a grid cell through each of eight points, as shown in Fig. 7.4. This differs from
Fig. 7.2, where uxes entered or left through each of four points. The locations of
variables and index values in Fig. 7.4 are the same as those in Fig. 7.2.
Substituting differential operators for difference operators in (7.28), multiplying
through by L
e
L, and interpolating variables in a manner consistent with kinetic
energy and enstrophy conservation give the momentum equation in the u-direction
as
215
j+1
j+1/2
j
j1/2
j1
j3/2
i3/2
i1
i1/2
i
i+1/2
i+1
a
v
u
u
u
u
u
u
u
u
u
v
v
v
v
v
v
v
v
j1
j1/2
j3/2
j
j+1/2
j+1
u
u
u
v
v
v
i3/2
j+3/2
j+3/2
i+1
i+3/2
i+3/2
i i1/2
i1
i+1/2
C
B
C
B
E
E
D
D
Figure 7.4 Location of uxes for nite-differencing the momentum
equation in the u-direction. The grid is the same as that in Fig. 7.2.
The gure shows exact locations of u, v, and
a
relative to lateral
boundaries.
Time difference term
u
i +12. j.k.t
=
(
a.th
LA)
i +12. j
(
a.t
LA)
i +12. j
u
i +12. j.k.th
+
h
(
a.t
LA)
i +12. j
(7.32)
Horizontal advection terms
B
i. j
u
i 12. j
+u
i +12. j
2
B
i +1. j
u
i +12. j
+u
i +32. j
2
+C
i +12. j 12
u
i +12. j 1
+u
i +12. j
2
C
i +12. j +12
u
i +12. j
+u
i +12. j +1
2
+D
i. j 12
u
i 12. j 1
+u
i +12. j
2
D
i +1. j +12
u
i +12. j
+u
i +32. j +1
2
+E
i +1. j 12
u
i +32. j 1
+u
i +12. j
2
E
i. j +12
u
i +12. j
+u
i 12. j +1
2
k.th
(7.33)
Vertical transport of horizontal momentum
+
1
L
k
(
a.t
LA
k12.t
u
k12.th
a.t
LA
k+12.t
u
k+12.th
)
i +12. j
(7.34)
216
Coriolis and spherical grid conversion terms
+
R
e
(L
e
L)
i +12. j
2
a.i. j
v
i. j 12
+v
i. j +12
2
f
j
R
e
cos
j
+
u
i 12. j
+u
i +12. j
2
sin
j
+
a.i +1. j
v
i +1. j 12
+v
i +1. j +12
2
f
j
R
e
cos
j
+
u
i +12. j
+u
i +32. j
2
sin
j
k.th
(7.35)
Pressure gradient terms
R
e
L
i +12. j
(+
i +1. j.k
+
i. j.k
)
a.i. j
+
a.i +1. j
2
+(
a.i +1. j

a.i. j
)
c
p.d
2
v.k
k+12
(P
k+12
P
k
) +
k12
(P
k
P
k12
)
L
k
i. j
+
v.k
k+12
(P
k+12
P
k
) +
k12
(P
k
P
k12
)
L
k
i +1. j
th
(7.36)
Eddy diffusion term
+
(
a.th
LA)
i +12. j
(
z

a
K
m
z
) u
i +12. j.k.th
(7.37)
All and some
a
values are evaluated at time t. All other terms are evaluated
at time t h. The interpolation of the second pressure gradient term in (7.36) is
differenced in a manner consistent with the differencing of geopotential, discussed
shortly.
The equations above require additional interpolations for certain terms. First,
column pressure multiplied by grid-cell area at a u-point is interpolated with
(
a
LA)
i +12. j
=
1
8
(
a
LA)
i. j +1
+(
a
LA)
i +1. j +1
+2[(
a
LA)
i. j
+(
a
LA)
i +1. j
]
+(
a
LA)
i. j 1
+(
a
LA)
i +1. j 1
(7.38)
where
LA =R
2
e
cos L
e
L (7.39)
is the horizontal area of a grid cell. Equation (7.38) relies on column pressures (at
times t or t h) and grid areas at
a
points. An interpolation similar to (7.38),
217
required for (7.34), is
(
a.t
LA
k12.t
)
i +12. j
=
1
8
(
a.t
LA
k12.t
)
i. j +1
+(
a.t
LA
k12.t
)
i +1. j +1
+2
(
a.t
LA
k12.t
)
i. j
+(
a.t
LA
k12.t
)
i +1. j
+(
a.t
LA
k12.t
)
i. j 1
+(
a.t
LA
k12.t
)
i +1. j 1
(7.40)
All values at time t on the right side of this equation are known from (7.14) and
(7.21) at the time the equation is evaluated.
A third set of interpolations, required for the uxes in (7.33) shown in Fig. 7.4,
is
B
i. j
=
1
12
[F
i 12. j 1
+ F
i +12. j 1
+2(F
i 12. j
+ F
i +12. j
)
+ F
i 12. j +1
+ F
i +12. j +1
] (7.41)
C
i +12. j 12
=
1
12
[G
i. j 32
+ G
i +1. j 32
+2(G
i. j 12
+ G
i +1. j 12
)
+G
i. j +12
+ G
i +1. j +12
] (7.42)
D
i. j +12
=
1
24
(G
i. j 12
+2G
i. j +12
+ G
i. j +32
+ F
i 12. j
+ F
i 12. j +1
+ F
i +12. j
+ F
i +12. j +1
) (7.43)
E
i. j +12
=
1
24
G
i. j 12
+2G
i. j +12
+ G
i. j +32
F
i 12. j
F
i 12. j +1
F
i +12. j
F
i +12. j +1
(7.44)
where F and G are the uxes dened in (7.15) and (7.16), respectively. Finally,
u-values at the bottom boundary of each layer, required for (7.34), may be inter-
polated vertically with
u
i +12. j.k+12.th
=
L
k+1
u
i +12. j.k.th
+L
k
u
i +12. j.k+1.th
L
k
+L
k+1
(7.45)
For nonnested, limited-area grids, special horizontal boundary conditions are
necessary for the momentum equation. A virtual row is added beyond the hori-
zontal boundary for each variable. For example, if i + 1 in Fig. 7.4 is the eastern
a
-point boundary for values of
a
, +, and , an additional row just outside the
boundary at
a
-point i + 2 is needed. A virtual row of u points at i + 52 out-
side the boundary is also needed. In the pressure gradient term of the momentum
equation, +
th
and
a.th
values in the virtual row outside the boundary are set to
+
t2h
and
a.t2h
values just inside the boundary (e.g., Lu 1994). In such a case,
(7.36) is replaced with
218
j+1
j+1/2
j
j1/2
j1
j3/2
i3/2
i1
i1/2
i
i+1/2
i+1
a

a
a
v
u
u
u
u
u
u
u
u
u
v
v
v
v
v
v
v
v
j1
j1/2
j3/2
j
j+1/2
j+1
u
u
u
v
v
v
i3/2
j+3/2
j+3/2
i+1
i+3/2
i+3/2
i i1/2
i1
i+1/2
Q
R
Q
T
T
S
S
R
Figure 7.5 Location of uxes for the nite-differencing of the momen-
tum equation in the v-direction. The grid is the same as that in
Fig. 7.2. The gure shows exact locations of u, v, and
a
relative to
lateral boundaries.
R
e
L
I+12. j
(+
I. j.k.t2h
+
I. j.k.th
)
a.I. j.th
+(
a.I. j.t2h
a.I. j.th
)
c
p.d
v.k
k+12
(P
k+12
P
k
) +
k12
(P
k
P
k12
)
L
k
I. j.th
(7.46)
This boundary condition dampens the effect of gravity waves, which rapidly create
instabilities when the boundary condition is absent.
Virtual boundary values of and u can be set to nearest inside boundary values
from the current time step or previous time step. Arakawa (1984) and Lu (1994)
contain a more rigorous treatment of these variables at the boundaries. In that
treatment, momentum uxes at the boundaries are derived so as to conserve mass
at the boundaries.
The momentumequation in the southnorth direction is differenced in a manner
similar to that in the westeast direction. Figure 7.5 shows the locations of uxes
for nite differencing this equation. The nite-difference form of the v-momentum
equation, given in (5.87), is
Time difference term
v
i. j +12.k.t
=
(
a.th
LA)
i. j +12
(
a.t
LA)
i. j +12
v
i. j +12.k.th
+
h
(
a.t
LA)
i. j +12
(7.47)
219
Horizontal advection terms
Q
i 12. j +12
v
i 1. j +12
+v
i. j +12
2
Q
i +12. j +12
v
i. j +12
+v
i +1. j +12
2
+R
i. j
v
i. j 12
+v
i. j +12
2
R
i. j +1
v
i. j +12
+v
i. j +32
2
+S
i 12. j
v
i 1. j 12
+v
i. j +12
2
S
i +12. j +1
v
i. j +12
+v
i +1. j +32
2
+T
i +12. j
v
i +1. j 12
+v
i. j +12
2
T
i 12. j +1
v
i. j +12
+v
i 1. j +32
2
k.th
(7.48)
Vertical transport of horizontal momentum
+
1
L
k
a.t
LA
k12.t
v
k12.th
a.t
LA
k+12.t
v
k+12.th
i. j +12
(7.49)
Coriolis and spherical grid conversion terms
R
e
(L
e
L)
i. j +12
2

a.i. j
u
i 12. j
+u
i +12. j
2
f
j
R
e
cos
j
+
u
i 12. j
+u
i +12. j
2
sin
j
+
a.i. j +1
u
i 12. j +1
+u
i +12. j +1
2
f
j +1
R
e
cos
j +1
+
u
i 12. j +1
+u
i +12. j +1
2
sin
j +1
k.th
(7.50)
R
e
(cos L
e
)
i. j +12
(+
i. j +1.k
+
i. j.k
)
a.i. j
+
a.i. j +1
2
+(
a.i. j +1
a.i. j
)
c
p.d
2
v.k
k+12
(P
k+12
P
k
) +
k12
(P
k
P
k12
)
L
k
i. j
+
v.k
k+12
(P
k+12
P
k
) +
k12
(P
k
P
k12
)
L
k
i. j +1
th
(7.51)
Eddy diffusion term
+(
a.th
LA)
i. j +12
(
z

a
K
m
z
) u
i. j +12.k.th
(7.52)
In the equations above, column pressure multiplied by the area of a grid cell at
a v-point is interpolated with
(
a
LA)
i. j +12
=
1
8
(
a
LA)
i +1. j
+(
a
LA)
i +1. j +1
+2[(
a
LA)
i. j
+(
a
LA)
i. j +1
]
+(
a
LA)
i 1. j
+(
a
LA)
i 1. j +1
(7.53)
220
7.6 The hydrostatic equation
In addition,
(
a.t
LA
k12.t
)
i. j +12
=
1
8
(
a.t
LA
k12.t
)
i +1. j
+(
a.t
LA
k12.t
)
i +1. j +1
+2[(
a.t
LA
k12.t
)
i. j
+(
a.t
LA
k12.t
)
i. j +1
]
+(
a.t
LA
k12.t
)
i 1. j
+(
a.t
LA
k12.t
)
i 1. j +1
(7.54)
Fluxes in (7.48) are found with
Q
i 12. j +12
=
1
12
[F
i 32. j
+ F
i 32. j +1
+2(F
i 12. j
+ F
i 12. j +1
)
+ F
i +12. j
+ F
i +12. j +1
] (7.55)
R
i. j
=
1
12
[G
i 1. j 12
+ G
i 1. j +12
+2(G
i. j 12
+ G
i. j +12
)
+G
i +1. j 12
+ G
i +1. j +12
] (7.56)
S
i +12. j
=
1
24
(G
i. j 12
+ G
i. j +12
+ G
i +1. j 12
+ G
i +1. j +12
+ F
i 12. j
+2F
i +12. j
+ F
i +32. j
) (7.57)
T
i +12. j
=
1
24
(G
i. j 12
+ G
i. j +12
+ G
i +1. j 12
+ G
i +1. j +12
F
i 12. j
2F
i +12. j
F
i +32. j
) (7.58)
Finally, the southnorth scalar velocity at the bottom of a layer can be interpolated
with
v
i. j +12.k+12.th
=
L
k+1
v
i. j +12.k.th
+L
k
v
i. j +12.k+1.th
L
k
+L
k+1
(7.59)
Boundary conditions for the v-equation are similar to those for the u-equation. For
instance, pressure-gradient boundary conditions in the v-equation are
R
e
(cos L
e
)
i.J +12
(+
i.J.k.t2h
+
i.J.k.th
)
a.i.J.th
+(
a.i.J.t2h
a.i.J.th
)
c
p.d
v.k
k+12
(P
k+12
P
k
) +
k12
(P
k
P
k12
)
L
k
i.J.th
(7.60)
7.6 THE HYDROSTATIC EQUATION
When a grid-cell dimension is greater than 3 km, the model atmosphere is generally
in hydrostatic equilibrium, and the geopotential used in the horizontal momentum
equations can be calculated diagnostically with the hydrostatic equation. From
(5.38), the hydrostatic equation was d+ = c
p.d
v
dP. A nite-difference form of
221
this equation giving the geopotential at the vertical midpoint of the bottom layer
at time t h is
+
i. j.N
L
.th
= +
i. j.N
L
+12
c
p.d
[
v.N
L
(P
N
L
P
N
L
+12
)]
i. j.th
(7.61)
(Arakawa and Suarez 1983) where P
N
L
+12
and P
N
L
are found from (7.9) and
(7.8), respectively. The geopotential at the surface, +
i. j.N
L
+12
, is the topographical
surface altitude multiplied by gravity. The geopotential at the bottom and vertical
midpoint of each layer above the surface are found from
+
i. j.k+12.th
= +
i. j.k+1.th
c
p.d
[
v.k+1
(P
k+12
P
k+1
)]
i. j.th
(7.62)
+
i. j.k.th
= +
i. j.k+12.th
c
p.d
[
v.k
(P
k
P
k+12
)]
i. j.th
(7.63)
respectively. The geopotential is solved for from the model surface to top. The
boundary conditions for the geopotential were discussed in Section 7.5.
7.7 ORDER OF CALCULATIONS
The equations in the preceding sections should be solved during a time step in the
following order: column pressure from (7.14), vertical scalar velocity from (7.21),
specic humidity and moist-air mass mixing ratios of gases and aerosol-particle
components from (7.24), potential virtual temperature from (7.27), geopotential
from(7.61)(7.63), westeast scalar velocities from(7.32)(7.37), and southnorth
scalar velocities from (7.47)(7.52). Quantities in the equations are used explicitly
(from time t h), except for all -values and specied
a
-values. Since most equa-
tions rely on and some
a
-values at time t, these parameters must be determined
rst and second, respectively.
7.8 TIME-STEPPING SCHEMES
In this chapter, prognostic equations were written as explicit, rst-order approxi-
mations in time (in forward Euler form). Since the forward Euler is uncondition-
ally unstable for any wind speed in the absence of eddy diffusion, as discussed in
Chapter 6, the forward Euler causes solutions to break down quickly. To increase
stability and the length of simulation, a better time-stepping scheme is needed.
Stability and accuracy can be improved by using the CrankNicolson time-stepping
scheme (Section 6.4.4.4), the Matsuno scheme alone (Section 6.4.4.6), the Matsuno
scheme combined with the leapfrog scheme (Section 6.4.4.5), the Heun scheme
(Section 6.4.4.7), the AdamsBashforth scheme (Section 6.4.4.8), or the fourth-
order RungeKutta scheme (Section 6.4.4.9). Figure 6.5 shows that the fourth-
order RungeKutta is generally the most accurate among these schemes when
222
7.8 Time-stepping schemes
applied to the one-dimensional advectiondiffusion equation. It is also the most
accurate and the most conservative when applied to the equations of atmospheric
dynamics (e.g., Ketean and Jacobson 2005a,b). The tradeoff is that it requires
more computer time per time step than the other schemes because it requires four
evaluations of all equations solved each time step rather than one or two evalua-
tions, in the case of the other schemes listed.
Due to computer time constraints, most atmospheric models use a simpler time-
stepping scheme. The Matsuno scheme is rst-order in time but conditionally stable.
This scheme, alone, can be used to solve the equations of atmospheric dynamics.
During a Matsuno time step, variable values at time t are rst estimated fromvalues
at time t h. If the time derivative of the moist-air mass mixing ratio of a species
is
q
t
= f (q) (7.64)
then q is estimated explicitly with
q
est
= q
th
+hf (q
th
) (7.65)
The estimate is then used to calculate the spatial derivative in a second equation
that also relies on information from one time step backward. The nal value of q
from the second equation is
q
t
= q
th
+hf (q
est
) (7.66)
Thus, with the Matsuno scheme, the right sides of all equations are evaluated twice
during each time step.
The time-derivative approximation can be improved to second order by using
the leapfrog scheme. Since the leapfrog scheme destabilizes over time when used
alone to solve nonlinear equations, it cannot be used alone. Thus, the Matsuno
scheme or another scheme must be inserted every 515 leapfrog steps to maintain
stability. Combining the Matsuno scheme, which is a rst-order approximation,
with the leapfrog scheme, which is a second-order approximation, gives a resulting
approximation between rst and second order.
With the leapfrog scheme, time derivatives leap over a time step in the spatial
derivative terms. The mixing ratio at the end of a leapfrog time step is
q
t+h
= q
th
+2hf (q
t
) (7.67)
which requires information from two time steps backward. The Matsuno scheme
is used not only to stabilize the leapfrog scheme but also to initialize a new simula-
tion or time interval. Figure 7.6 shows a sequence of dynamics calculations when
Matsuno and leapfrog schemes are used together. In Chapter 6, a time step was
dened as an increment in time for a given algorithm, and a time interval was
dened as the period during which several time steps of a process are completed
without interference by another process. In Fig. 7.6, M1, L2, L3, . . . , L6 are time
223
L2
L4 L6
L3 L5 M1
L2
L4 L6
L3 L5 M1
Figure 7.6 Example of a dynamical time-stepping scheme that uses a Mat-
suno (M) and a leapfrog (L) scheme. Two interruptions with a Matsuno
step are shown one to start the simulation and one after several leapfrog
steps. The number next to each letter indicates the order in which the step
is taken during a sequence.
steps, and each set of M1, . . . , L6 is a time interval. A typical dynamic time step
in a global model is 60300 s. That in a mesoscale model is 330 s.
7.9 SUMMARY
In this chapter, nite-difference solutions to the equations of atmospheric dynam-
ics, including the continuity equation for air, the species continuity equation, the
thermodynamic energy equation, the horizontal momentum equations, and the
hydrostatic equation, were given. The thermodynamic energy equation is used to
calculate the potential virtual temperature. The horizontal momentum equations
are used to obtain horizontal scalar velocities. When the atmosphere is in hydro-
static equilibrium, the continuity equation for air is used to solve for vertical scalar
velocity and column pressure, and the hydrostatic equation is used to solve for
geopotential. A common time-stepping scheme in dynamical models is the combi-
nation of a Matsuno and leapfrog scheme.
7.10 PROBLEMS
7.1 For the same time step and grid spacing, should the model accuracy increase
or decrease when Matsuno but no leapfrog steps are taken versus when both
Matsuno and leapfrog steps are taken? Why? How is the computer time
affected in each case?
7.2 (a) Assume a grid cell, centered at = 45
N, has dimensions L
e
= 5
,
L = 4
, and L = 0.05. Calculate q at the center of the cell after one

5-min time step if u
1
= +3 m s
1
,
a,1
= 745 hPa, q
1
= 0.0062 kg kg
1
;
u
2
= +1 m s
1
,
a,2
= 752 hPa, q
2
= 0.0047 kg kg
1
; v
3
= + 2 m s
1
,
a,3
= 754 hPa, q
3
= 0.0064 kg kg
1
; v
4
= + 1 m s
1
,
a,4
= 746 hPa,
q
4
= 0.0054 kg kg
1
,
5
= 4.58 10
6
s
1
, q
5
= 0.005 kg kg
1
; and

6
= 6.11 10
6
s
1
, q
6
= 0.006 kg kg
1
, where subscripts 1, . . . , 6
denote the west, east, south, north, lower, and upper boundaries, respec-
tively. Average the four
a
-values andsix q-values to obtaincenter values.
Neglect eddy diffusion and external sources and sinks. Assume the air
is dry and the
a
-values stay constant for the time step.
(b) For the conditions of part (a), calculate
v
at the center of the cell after
30 minutes if the west, east, south, north, lower, and upper grid-cell
224
7.12 Modeling project
boundary values are
v
= 297, 298, 300, 299, 298.3, and 298.7 K, respec-
tively. Average the six boundary values to obtain an initial center value
of
v
.
7.3 Set up a 20-layer (21-boundary) sigma-pressure-coordinate vertical model
grid between p
a,top
= 250 hPa and p
a,surf
= 1000 hPa at sea level. Assume
each layer has equal sigma thickness, and assume T = 288 K at the sur-
face. Assume also that temperature decreases with increasing altitude at
6.5 K km
1
. Calculate pressure at the center of each layer with (7.10). Plot
pressure versus altitude.
7.12 MODELING PROJECT
The purpose of this project is to develop a basic regional- or global-scale model
that solves the equations of atmospheric dynamics, except for the water-vapor
continuity equation. Diabatic energy sources and sinks and eddy diffusion are
ignored. Winds are driven primarily by pressure gradients.
(a) Choose a southwest-corner latitude and longitude anywhere on Earth away
from the poles.
(b) Set a model top pressure everywhere of 250 hPa (near the tropopause).
(c) Set NLAT, NLONG, and NVERT, which denote the numbers of latitudi-
nal, longitudinal, and vertical grid centers in the model. Note that 1, . . . ,
NLAT, 1, . . . , NLONG, and 1, . . . , NVERT increase from south to north,
west to east, and top to bottom, respectively. NLAT + 1, NLONG + 1, and
NVERT + 1 equal the numbers of boundaries in the respective directions.
Allow parameters to be variable, but set them initially to NLAT = 40,
NLONG = 50, and NVERT = 15.
(d) Use spherical horizontal coordinates and the sigma-pressure vertical coor-
dinate. Select L
e
=0.05
and L =0.05
. Select values of L for each layer

using the method discussed in this chapter. Ensure that the bottom layer is
at least 150 m thick (1 hPa 10 m).
(e) Initialize surface pressure by creating a small hill of pressure in the center of
the grid superimposed on background pressure. First, select a background
surface pressure ( p
a.base
) of 1000 hPa and a peak incremental surface pressure
(Lp
a.peak
) of 3 hPa. Calculate the surface pressure at the horizontal center of
each grid cell i, j with the Gaussian distribution,
p
a.surf.i. j
= p
a.base
+Lp
a.peak
exp
R
e
cos
i. j
+
c
2
i. j

c
2
2

[R
e
(
i. j

c
)]
2
2
225
where
c
and
c
are the longitude and latitude (radians), respectively, at the
center of your grid. Calculate the initial column pressure at the horizontal
center of each grid cell with
i. j
p
a.surf.i. j
p
top
.
(f) Use (7.6) to calculate pressure at the horizontal midpoint/vertical bound-
ary of each cell. Use (7.10) to calculate pressure at the horizontal mid-
point/vertical midpoint of each cell from horizontal midpoint/vertical
boundary values.
(g) Initialize temperatures at the horizontal midpoint/vertical midpoint of
each cell in one corner column of the model. Estimate temperatures from
Appendix Table B.1 given the pressures from part (f).
(h) Set the temperature at the horizontal midpoint/vertical midpoint of each
cell in all other columns equal to that in the initialized column. Calculate
potential virtual temperature at the horizontal midpoint/vertical midpoint
of each cell. Assume the air is dry. Interpolate potential virtual temperature
to the horizontal midpoint/vertical boundary of each cell with (7.11). No
values are needed for the model top or bottom boundaries (k = 12 or N
L
+
12).
(i) Use the equation of state to calculate the density at the horizontal mid-
point/vertical midpoint of each cell.
(j) Use (7.14) to solve for the column pressure at the horizontal midpoint of
each column in the model.
(k) Use (7.21) to solve for vertical velocities at the horizontal midpoint/vertical
boundary of each cell.
(l) Use (7.27) to solve for the potential virtual temperature at the horizontal
midpoint/vertical midpoint of each cell. Ignore diabatic sources and sinks
and ignore eddy diffusion.
(m) Solve for the geopotential in each cell with (7.61)(7.63). Assume the surface
geopotential is zero.
(n) Use (7.32)(7.37) and (7.47)(7.52) to solve for u and v at locations shown
in Figs. 7.4 and 7.5, respectively. Ignore eddy diffusion. Assume u, v, and
are initially zero everywhere.
(o) Use a Matsuno or Matsuno-plus-leapfrog time-stepping scheme instead
of the forward Euler scheme to advance all prognostic equations for this
problem. Recalculate all diagnostic variables during each Matsuno and/or
leapfrog step.
(p) Debug, and run simulations. Change conditions by changing the location
and/or magnitude of peak surface pressure.
(q) Test the effect of adding topography by changing the geopotential at the
surface.
(r) Plot horizontal and/or vertical elds of temperature, pressure, andvelocities.
(s) Instead of using horizontal limited-area boundary conditions as described in
the text, it may be more convenient, for this project, to use periodic boundary
conditions. To accomplish this, set up the limited-area model grid so that it
is symmetric in the southnorth direction about the Equator, allow ow out
of the north boundary edge to enter the south boundary edge and vice versa,
and allow ow out of the east boundary edge to enter the west boundary
226
edge and vice versa. For this example case, the Coriolis parameter, f, may
be set to zero, otherwise it will have opposite signs on the north and south
boundaries, which connect to each other. Allowing f to vary, however, does
ensure enstrophy and energy conservation. Although not necessarily realis-
tic, this setup will simplify the coding, and allow for useful sensitivities of
the numerical properties of the code.
227
8
Boundary-layer and surface processes
T
he boundary layer is the region of the atmosphere between the Earths sur-
face and 5003000 m height that is inuenced substantially by energy and
moisture from the surface. The bottom ten percent of the boundary layer is
the surface layer. Some parameters that affect the surface layer and boundary
layer are ground temperature, soil moisture, and turbulent uxes. In this chap-
ter, expressions for turbulent uxes of momentum, energy, and moisture are
given in terms of bulk aerodynamic formulae, MoninObukhov similarity theory,
K-theory, and higher-order parameterizations. From these equations, eddy diffu-
sion coefcients for momentum and energy are derived. Analytical equations for
the vertical prole of wind speed, potential virtual temperature, and moisture in
the surface layer are also provided. Prognostic equations for surface temperature
and soil moisture are then given. Types of surfaces treated include soil, water,
roads, rooftops, sea ice, vegetation over soil, and snowover all surfaces. Boundary-
layer and surface processes affect weather and air pollution from urban to global
scales.
8.1 TURBULENT FLUXES OF MOMENTUM, ENERGY,
AND MOISTURE
Turbulence is due to wind shear (mechanical turbulence) and buoyancy (thermal
turbulence). Turbulence mixes gradients of momentum, energy, moisture, gases,
and particles vertically and horizontally. In a model, the degree of vertical mixing
due to turbulence can be quantied with a turbulent ux term. In the case of
vertical mixing of horizontal momentum, the term is a function of the kinematic
vertical turbulent momentum ux, w
and w
(m
2
s
1
) (Section 4.2.6). In the
case of vertical mixing of energy, it is a function of the kinematic vertical turbulent
sensible-heat ux, w
v
(m K s
1
) (Section 3.4). In the case of vertical mixing of
moisture, it is a function of the kinematic vertical moisture ux, w
v
(m kg s
1
kg
1
). Analogous uxes can be developed for other gases and particles.
Kinematic turbulent uxes of momentum are negatively proportional to
Reynolds stresses. A stress is a force per unit area that causes a body to deform.
A Reynolds stress, which arises when a uid undergoes turbulent motion, causes
a parcel of air to deform. Suppose an air parcel uctuates randomly in time due
228
8.1 Turbulent fluxes of momentum, energy, and moisture
Eddy
mixing
x
y
z
Force arising
from w'u'
Figure 8.1 Deformation of a cubic
air parcel caused by a kinematic
vertical momentum ux w
, as
described in the text. Adapted from
Stull (1988).
to mechanical shear and buoyancy. Precise scalar velocities, w and u, in the par-
cel have mean components w and u and eddy components w
and u
, respectively.
A w
-velocity has the effect of mixing u
-velocities in the z-direction. The vertical

mixing of the u
-velocity exerts a force in the x-direction over an area normal to

the z-direction. The effect of the force per unit area, or stress, is to induce a drag
on u and to cause the air parcel to deform, as shown in Fig. 8.1.
In the example above, the scalar component of Reynolds stress in the x-direction
along a plane normal to the z-direction is
zx
=
a
w
(8.1)
(kg m
1
s
2
or N m
2
), where
a
is air density (kg m
3
). This stress results from
the vertical transport of a westeast gradient of momentum. Since w
mixes u
in
the same way that u
mixes w
, we have u
= w
,
zx
=
xz
, and
xz
=
a
u
.
The Reynolds stress in the y-direction along a plane normal to the z-direction is
zy
=
a
w
. Both
zx
and
zy
have the effect of transporting gradients of horizon-
tal momentum vertically. Combining the two gives the magnitude of the vertical
turbulent ux of horizontal momentum as
|
z
| =
a
(w
)
2
+(w
)
2
1
/
2
(8.2)
(kg m
1
s
2
). From (8.1), the kinematic vertical turbulent uxes of westeast and
southnorth momentum are
w
zx
a
w
zy
a
(8.3)
Fluxes such as these are needed in the turbulent-ux divergence termof the momen-
tum equation given in (4.66). The uxes can be measured, but in a model, they
need to be parameterized, as discussed in Section 8.4.
229
Example 8.1
Reynolds stresses are usually much larger than viscous stresses. From (4.56),
the shearing stress in the x-direction over the xy plane due to wind shear
in the z-direction was
zx
=
a
uz. If T = 288 K and u = 5 m s
1
at z = 10 m,
then
a
= 1.792 10
5
kg m
1
s
1
and uz = 0.5 s
1
. The resulting shearing
stress due to viscosity is
zx
= 8.96 10
6
kg m
1
s
2
.
A typical value of n
is 0.4 m
2
s
2
. For a surface air density of
a
=
1.25 kg m
3
, (8.1) predicts a Reynolds stress due to turbulence of
zx
=
0.5 kg m
1
s
2
.
Gradients of energy, like momentum, are transferred vertically by turbulence.
The vertical turbulent sensible-heat ux is dened as
H
f
=
a
c
p.d
w
v
(8.4)
(W m
2
) where c
p.d
is the specic heat of dry air at constant pressure (J kg
1
K
1
),
v
is the turbulent uctuation of potential virtual temperature (K), and w
v
(m K s
1
) is the kinematic vertical turbulent sensible-heat ux. Rewriting (8.4)
gives
w
v
=
H
f
a
c
p.d
(8.5)
which is needed to calculate the effect of turbulence on mixing of energy gradients in
the thermodynamic energy equation of (3.72). The kinematic ux can be calculated
fromobserved quantities of
v
and w
or parameterized, as discussed in Section 8.4.

Gradients of water vapor are also transferred vertically by turbulence. The ver-
tical turbulent ux of water vapor is
E
f
=
a
w
v
(8.6)
(kg m
2
s
1
) where q
v
is the turbulent uctuation of water-vapor specic humidity
(kg kg
1
), and w
v
(m kg s
1
kg
1
) is the kinematic vertical turbulent moisture
ux. Rewriting (8.6) gives
w
v
=
E
f
a
(8.7)
which is needed to calculate the effect of turbulence on mixing of moisture gradients
in the water vapor continuity equation of (3.49).
8.2 FRICTION WIND SPEED
When horizontal winds ow over roughness elements protruding from a surface,
drag slows the wind near the surface relative to the wind aloft, creating vertical wind
shear. Wind shear produces eddies that exchange momentum, energy, gases, and
aerosol particles vertically. The greater the height that roughness elements protrude
from a surface and the greater the horizontal wind speed, the greater the resulting
230
8.3 Surface roughness lengths
wind shear and vertical ux of horizontal momentum. A scaling parameter that
provides a measure of the vertical ux of horizontal momentumin the surface layer
(denoted by the subscript s) is the friction wind speed,
u
(w
)
2
s
+(w
)
2
s
14
=
|
z
|
12
s
(8.8)
The greater the friction wind speed, the greater mechanical turbulence, and
the faster that momentum, energy, and pollutants from aloft mix down to the
surface and vice versa. Typical roughness elements at the surface include rocks,
trees, buildings, grass, and sand. The friction wind speed can be parameterized or
found from eld experiments in which u
, v
, and w
are measured.
Example 8.2
If two measurements of westeast and vertical scalar velocity yield u
1
=
10 m s
1
, w
1
= 0.1 m s
1
, and u
2
= 6 m s
1
, w
2
= 0.2 m s
1
, respectively,
estimate u
. Assume the southnorth scalar velocity is zero.

SOLUTION
u = (u
1
+u
2
)2 = 8 m s
1
n = (n
1
+n
2
)2 = 0.15 m s
1
u
1
= u
1
u = 2 ms
1
n
1
= n
1
n = 0.05 m s
1
u
2
= u
2
u = 2 m s
1
n
2
= n
2
n = 0.05 m s
1
Therefore,
n
= (n
1
u
1
+n
2
u
2
)2 = 0.1m
2
s
2
u
= (n
)
12
s
= 0.32 m s
1
8.3 SURFACE ROUGHNESS LENGTHS
Three variables used frequently in boundary-layer parameterizations are the sur-
face roughness lengths for momentum, energy, and moisture. The surface roughness
length for momentum (z
0,m
), or aerodynamic roughness, is the height above a sur-
face at which the logarithmic prole of wind speed versus altitude extrapolates
to zero wind speed. It gives a measure of vertical turbulence that occurs when a
horizontal wind ows over a rough surface. The greater z
0,m
, the greater the mag-
nitude of turbulence that arises when wind passes over a roughness element. For a
perfectly smooth surface, the roughness length is zero, and mechanical turbulence
is minimized. For other surfaces, it is sometimes approximated as 130th the height
of the average roughness element protruding from the surface. For surfaces with
sparsely placed roughness elements, z
0,m
is the height above the base of the rough-
ness elements (Brutsaert 1991). For densely placed roughness elements of average
231
ln z
0,m
|v
h
|
0
0
l
n

z
Figure 8.2 The surface roughness
length for momentum is calculated
by taking wind speed measurements
at several heights at a given loca-
tion when the wind is strong. The
data are plotted as the natural log
of height versus wind speed. Wind
speeds from higher altitudes on the
plot are then extrapolated until a
zero wind speed intersects the alti-
tude axis. The altitude at which
this occurs is the surface roughness
length for momentum. Points in the
diagram are measurements.
height h
c
, z
0,m
is the height above a displacement height (d
c
). Displacement heights
usually lie between 0 and h
c
(Section 8.4.2.8).
Figure 8.2 describes a method of estimating z
0,m
from observed wind speed
proles. Parameterizations of z
0,m
from eld data have also been developed. For
smooth surfaces, such as over a smooth ocean with low wind speeds, one param-
eterization is
z
0.m
0.11
a
u
= 0.11

a
a
u
(8.9)
(e.g., Hinze 1975; Garratt 1992), where z
0.m
is in m,
a
=
a
a
is the kinematic
viscosity of air (m
2
s
1
),
a
is the dynamic viscosity of air (kg m
1
s
2
) from (4.54),
and
a
is air density (kg m
3
). Over a rough ocean with high wind speeds,
z
0.m

c
u
2
g
(8.10)
which is the Charnock relation (Charnock 1955), where
c
0.016 is the
Charnock constant (e.g., Garratt 1992). Over urban areas containing structures,
z
0.m
0.5
h
0
S
0
A
0
(8.11)
(Lettau 1969; Petersen 1997), where 0.5 is the average drag coefcient over the
structures, h
0
(m) is the average structure height, A
0
(m
2
) is the area over which
z
0,m
is estimated, and S
0
(m
2
) is the silhouette area, measured in a vertical plane
normal to the mean wind, of all structures in surface area A
0
.
232
8.3 Surface roughness lengths
A canopy is the vegetation cover provided by a group of plants or trees. An
expression for the surface roughness length (m) over a vegetation canopy is
z
0.m
= h
c
(1 0.91e
0.0075L
T
) (8.12)
(Sellers et al. 1996), where h
c
is the canopy height (height in meters above the
ground of the top of a vegetation canopy) and L
T
is the one-sided leaf area index
(square meters of leaf surfaces per square meter of underlying ground). The one-
sided leaf area index measures a canopy area density and is calculated by integrating
the foliage area density, which is the area of plant surface per unit volume of air,
from the ground to the canopy height. The upper limit of L
T
may be 7 f
v
, where f
v
is the fraction of land surface covered by vegetation. Deardorff (1978) argued that,
at higher values of L
T
, insufcient light is available to support further growth in a
canopy. Table 8.1 gives z
0.m
for several surfaces, including those with vegetation.
Canopy heights and upper limits to leaf area indices (L
T.max
) are given where
applicable.
Example 8.3
If T = 288 K, p
d
= 1013 hPa, u
= 0.165 m s
1
, and the air is dry, estimate z
0.m
over a smooth ocean. Find z
0.m
over a rough sea when u
= 0.5 m s
1
.
SOLUTION
From (4.54),
a
= 0.0000179 kg m
1
s
1
; from (2.23),
a
=
d
= 1.225 kg m
3
;
and from (4.55),
a
= 0.0000146 m
2
s
1
. Thus, from (8.9), z
0.m
1 10
5
m
over a smooth sea. From (8.10), z
0.m
0.0004 m over a rough sea.
The surface roughness length for momentum characterizes the ability of surface
elements to absorb momentum. Surface roughness lengths for energy and water
vapor characterize the ability of elements protruding from the surface to absorb
energy and moisture, relative to their ability to absorb momentum (Kaimal and
Finnigan 1994). Both lengths are integration constants, used to derive vertical
proles of potential virtual temperature and specic humidity, respectively, in the
surface layer. Brutsaert (1991) gives several parameterizations of z
0,h
and z
0,v
based
on measurements over different terrain. More general expressions for the energy
and moisture roughness lengths are
z
0.h
=
D
h
ku
z
0.v
=
D
v
ku
(8.13)
respectively (Garratt and Hicks 1973), where D
h
is the molecular thermal diffusion
coefcient (molecular thermal diffusivity), D
v
is the molecular diffusion coefcient
of water vapor, and k is the von K arm an constant. The von K arm an constant has
a value between 0.35 and 0.43 (e.g., Hogstrom 1988). A value of 0.40 is assumed
233
Table 8.1 Aerodynamic roughnesses (z
0
,
m
), structure or canopy heights (h
c
), displacement
heights (d
c
), and maximum one-sided leaf area indices (L
T,max
) for several surfaces
L
T,max
Surface type z
0, m
(m) h
c
(m) d
c
(m) (m
2
m
2
) Reference
Smooth sea 0.00001 (8.9)
Rough sea 0.0000150.0015 (8.10)
Ice 0.00001 Oke (1978)
Snow 0.000050.0001 Oke (1978)
Level desert 0.0003 Sehmel (1980)
Short grass 0.0030.01 0.020.1 -0.075 Oke (1978)
Long grass 0.040.1 0.251.0 0.190.75 Oke (1978)
Savannah 0.4 8 4.8 Garratt (1992)
Agricultural crops 0.040.2 0.42 0.271.3 Oke (1978)
Orchard 0.51.0 510 3.36.7 Oke (1978)
Coniferous forest 0.283.9 10.427.5 6.325.3 Jarvis et al.
(1976)
Tropical forest 2.2 35 29.8 Shuttleworth
(1989)
Broadleaf
evergreen forest
4.8 35 26.3 7 Sellers et al.
(1996)
Broadleaf
deciduous trees
2.7 20 15 7 Sellers et al.
(1996)
Broad- and
needleleaf trees
2.8 20 15 7.5 Sellers et al.
(1996)
Needleleaf
evergreen trees
2.4 17 12.8 8 Sellers et al.
(1996)
Needleleaf
deciduous trees
2.4 17 12.8 8 Sellers et al.
(1996)
Short
vegetation/C4
grassland
0.12 1 0.75 5 Sellers et al.
(1996)
Broadleaf shrubs
w/bare soil
0.06 0.5 0.38 5 Sellers et al.
(1996)
Dwarf trees and
shrubs
0.07 0.6 0.45 5 Sellers et al.
(1996)
Agriculture/C3
grassland
0.12 1 0.75 5 Sellers et al.
(1996)
2500-m
2
lot with a
building 8 m
high and 160-m
2
silhouette
0.26 8 (8.11)
25 000-m
2
lot with
a building 80 m
high and
3200-m
2
silhouette
5.1 80 (8.11)
Displacement heights in most cases were calculated as 0.75 h
c
.
234
8.4 Parameterizations of kinematic turbulent fluxes
here. Expressions for D
h
and D
v
(m
2
s
1
) are
D
h
=

a
a
c
p.m
D
v
= 2.11 10
5
T
273.15 K
1.94
1013.25 hPa
p
a
(8.14)
respectively (Pruppacher and Klett 1997), where
a
is the thermal conductivity of
moist air (J m
1
s
1
K
1
),
a
is the density of air (kg m
3
), c
p.m
is the specic heat
of moist air at constant pressure (J kg
1
K
1
), T is temperature (K), and p
a
is air
pressure (hPa). For very smooth surfaces, z
0,m
is slightly less than z
0,h
and z
0,v
. For
rough surfaces, z
0,m
is up to 10
5
times larger than z
0,h
and z
0,v
(Brutsaert 1991).
A typical ratio is about 100 (Garratt and Hicks 1973). For rough surfaces, surface
elements are more efcient at absorbing momentum than they are at absorbing
energy or water vapor.
Example 8.4
Assuming u
= 0.4 m s
1
, p
d
= 1013 hPa, and the air is dry, calculate the
surface roughness lengths for energy and water vapor when T = 298 K.
SOLUTION
At T = 298 K,
a

d
= 0.0256 J m
1
s
1
K
1
from (2.5) and
a

d
= 1.18 kg
m
3
from (2.23). Thus, D
h
= 2.16 10
5
m
2
s
1
and D
:
= 2.50 10
5
m
2
s
1
from (8.14). Substituting these values into (8.13) gives z
0.h
= 0.000135 m and
z
0.:
= 0.000156 m, respectively.
8.4 PARAMETERIZATIONS OF KINEMATIC
TURBULENT FLUXES
Most atmospheric turbulence is produced by mechanical shear and buoyancy in
the form of large eddies. Large eddies are the size of the boundary layer, which is
on the order of hundreds of meters to three thousand meters in diameter. This size
range of eddies is called the energy-containing subrange. Eddies carry turbulent
kinetic energy (TKE), which is the mean kinetic energy per unit mass associated
with eddies in turbulent ow. Large eddies shrink over time due to dissipation,
which is the conversion of turbulence into heat by molecular viscosity. As a large
eddy shrinks to a small eddy, TKE stored in the eddy decreases, ultimately to zero,
in a process called inertial cascade. Thus, TKE is not conserved over time. The rate
of TKE transfer to small eddies is generally proportional to the dissipation rate.
The dissipation rate is greatest for the smallest eddies, which are on the order of
millimeters in diameter. Medium-sized eddies, which are created by dissipation of
large eddies but equally destroyed by dissipation to small eddies, are in the inertial
subrange of eddy size.
Atmospheric models with horizontal grid size greater than a fewhundred meters
on edge do not resolve the large eddies in turbulence. These models need to
235
parameterize kinematic turbulent uxes of momentum, sensible heat, and moisture,
dened in (8.3), (8.5), and (8.7), respectively, used in the momentum, thermody-
namic energy, and water vapor continuity equations. Such models are referred to
as Reynolds-averaged models (Section 3.3.1).
In the other extreme are models that resolve eddies of all sizes. The smallest
eddies in the atmosphere must be larger than the Kolmogorov scale,
k
=
3
a
14
(8.15)
(m) which is the length scale of turbulent motion below which the effects of molec-
ular viscosity are nonnegligible. In this equation,
a
is the kinematic viscosity of
air from (4.55) (m
2
s
1
) and
d
is the dissipation rate of turbulent kinetic energy
(m
2
s
3
), which is the rate of conversion of turbulence into heat by molecular vis-
cosity. A typical cloud-free dissipation rate is 0.0005 m
2
s
3
. Example 8.5 shows
that this dissipation rate gives a Kolmogorov scale of about 1.6 mm. Thus, the
smallest eddies in the atmosphere are a few millimeters in diameter. Models that
have dimension on the order of the Kolmogorov scale resolve all eddies and are
called direct numerical simulation (DNS) models. Since these models resolve tur-
bulence, they do not parameterize it. DNS models generally solve the unsteady,
incompressible equations of atmosphere dynamics (e.g., Kim et al. 1987; Joslin
et al. 1993; Coleman 1999).
Example 8.5
Calculate a typical Kolmogorov scale in the cloud-free atmosphere when
T = 288 K and p
d
= 1013 hPa.
SOLUTION
From Example 8.3,
a
= 0.0000146 m
2
s
1
at the temperature and pressure
given. A typical free-atmosphere turbulent dissipation rate is
d
= 0.0005 m
2
s
3
. Thus, from (8.15), the Kolmogorov scale is about 1.6 mm.
Models that have resolution between a few hundred meters and a few meters
(the scale of the inertial subrange) resolve large eddies but not small ones. These
models are referred to as large-eddy simulation (LES) models (e.g., Deardorff 1972;
Moeng 1984). They resolve turbulence on the large scales but need to parameterize
it on the small scales.
Reynolds-averaged models must parameterize kinematic turbulent uxes. Ide-
ally, these equations are solved prognostically. However, prognostic equations for
kinematic turbulent uxes contain more unknowns than equations available to
solve for them. When a new equation is written for an unknown, the equation cre-
ates more unknowns forcing at least some terms in the equation to be parameterized
as a function of known or derived variables. The inability to close equations for
236
turbulent ux terms without parameterizing some terms is called the problem of
closure. The order of closure is the highest order of a prognostic equation retained
in a parameterization. If, for example, a kinematic turbulent ux equation is prog-
nostic to rst order, and all higher-order terms are parameterized, the equation is
a rst-order closure equation. Stull (1988) contains more details about the closure
problem.
In the surface layer, kinematic turbulent uxes are often parameterized with bulk
aerodynamic formulae or MoninObukhov similarity theory. Both methods are
zero-order closure techniques in that the resulting equations are fully parameterized
and have no prognostic parts. Gradient transport theory (K-theory) is generally a
rst-order closure technique. K-theory assumes that the kinematic turbulent ux of
a quantity is negatively proportional to a constant diffusion coefcient multiplied
by the gradient of the mean value of the quantity. When bulk aerodynamic formulae
or scaling parameters from similarity theory are used to calculate eddy diffusion
coefcients in a K-theory expression for kinematic turbulent ux, the result is also
a rst-order closure equation. When the diffusion coefcient in K-theory is solved
as a function of several other variables, the resulting K-theory expressions can be
made second order or higher. In the next subsections, closure techniques of several
orders are described.
8.4.1 Bulk aerodynamic formulae
Bulk aerodynamic formulae are equations for surface-layer kinematic vertical tur-
bulent uxes that assume a constant drag coefcient. A constant drag coefcient
is most applicable when the boundary layer is well mixed, the surface layer is thin,
and wind speeds, potential virtual temperature, and moisture do not change signif-
icantly with height above the surface layer. Below, parameterizations of kinematic
vertical turbulent uxes and eddy diffusion coefcients from bulk aerodynamic
theory are given for momentum, energy, and moisture.
8.4.1.1 Momentum uxes
In the absence of wind shear above the thin surface layer, the vertical turbulent
transfer of horizontal momentumis affected primarily by skin drag, formdrag, and
wave drag. Skin drag is the near-surface drag that results from molecular diffusion
of momentum across the surfaceair interface. Form drag is the near-surface drag
resulting from turbulence and vertical momentum transfer that occurs when winds
hit large obstacles, such as rocks or trees. Wave drag is the near-surface drag that
results from vertical transfer of momentum by gravity waves, which propagate
vertically and horizontally, as discussed in Chapter 4. These three types of drag are
embodied in the dimensionless coefcient of drag C
D
, which is used in expressions
for bulk aerodynamic kinematic turbulent momentum uxes in the surface layer,
(w
)
s
= C
D
| v
h
(z
r
)|[ u(z
r
) u(z
0.m
)] (8.16)
(w
)
s
= C
D
| v
h
(z
r
)|[ v(z
r
) v(z
0.m
)] (8.17)
237
In these equations, u(z
0.m
) = v(z
0.m
) = 0 are the mean scalar velocities at height
z
0.m
, z
r
is a reference height above the surface (usually 10 m), and | v
h
(z
r
)| is the
mean horizontal wind speed at the reference height. The coefcient of drag, which
is evaluated at the reference height, ranges from 0.001 for smooth surfaces to 0.02
for rough surfaces. Over the ocean a typical value is 0.0015. In tropical storms and
hurricanes, C
D
varies with wind speed over the ocean, since increased wind speeds
increase ocean surface roughness (e.g., Krishnamurti et al. 1998).
From (4.67), the kinematic vertical turbulent uxes of momentum in the surface
layer from K-theory were dened as
(w
)
s
= K
m.zx
u
z
(w
)
s
= K
m.zy
v
z
(8.18)
The nite-difference form of the scalar-velocity gradient in each case is
u
z
=
u(z
r
) u(z
0.m
)
z
r
z
0.m
v
z
=
v(z
r
) v(z
0.m
)
z
r
z
0.m
(8.19)
respectively. Substituting (8.19), (8.16), and (8.17) into (8.18) and solving give the
K-theory eddy diffusion coefcient for momentum (m
2
s
1
) in the surface layer in
terms of bulk aerodynamic formulae as
K
m.zx
= K
m.zy
C
D
| v
h
(z
r
)| (z
r
z
0.m
) (8.20)
Example 8.6
If C
D
=0.02 (rough surface), u =10 ms
1
, and v =0 ms
1
at z =10 m, estimate
K
m.zx
.
SOLUTION
Assuming z
0.m
is negligible relative to z, we have K
m.zx
= 2 m
2
s
1
from (8.20).
8.4.1.2 Energy uxes
From bulk aerodynamic formulae the kinematic vertical turbulent sensible-heat
ux in the surface layer is
(w
v
)
s
= C
H
| v
h
(z
r
)| [
v
(z
r
)

v
(z
0.h
)] (8.21)
where C
H
is the bulk heat-transfer coefcient or Stanton number (dimensionless),
and

v
(z
0.h
) is the mean potential virtual temperature at height z
0.h
. In a model,

v
at the ground is often used instead of

v
(z
0.h
) for convenience. The ground value
of

v
can be estimated with a soil model, as discussed in Section 8.6. C
H
depends
on the molecular thermal diffusivity across the surfaceair interface (skin drag). Its
value ranges from 0.001 for smooth surfaces to 0.02 for rough surfaces. C
H
differs
from C
D
because momentum is affected by skin, form, and wave drag while energy
is affected primarily by skin drag.
238
Substituting (8.21) into (8.4) gives the vertical turbulent sensible heat ux near
the surface as
H
f

a
c
p.d
C
H
| v
h
(z
r
)| [
v
(z
0.h
)

v
(z
r
)] (8.22)
(W m
2
). This equation states that energy is transferred upward when

v
at z
0.h
is
greater than that at z
r
.
From (3.73), the kinematic vertical turbulent sensible-heat ux in the surface
layer from K-theory was dened as
(w
v
)
s
= K
h.zz
v
z
(8.23)
A nite-difference form of the gradient of potential virtual temperature is
v
z
=
v
(z
r
)

v
(z
0.h
)
z
r
z
0.h
(8.24)
Substituting (8.21) and (8.24) into (8.23) gives the K-theory vertical eddy diffusion
coefcient for energy (m
2
s
1
) in the surface layer in terms of bulk aerodynamic
formulae as
K
h.zz
C
H
| v
h
(z
r
)|(z
r
z
0.h
) (8.25)
8.4.1.3 Moisture uxes
A bulk aerodynamic equation for the kinematic vertical turbulent water-vapor ux
is
(w
v
)
s
= C
E
| v
h
(z
r
)| [ q
v
(z
r
) q
v
(z
0.v
)] (8.26)
where C
E
(dimensionless) is the bulk transfer coefcient for water vapor and
q
v
(z
0.v
) is the mean specic humidity at height z
0.v
. In a model, the specic
humidity in the top molecular soil or water surface layers is often used instead
of q
v
(z
0.v
). C
E
is usually set equal to C
H
. Kinematic vertical turbulent uxes of
gases other than water vapor and aerosol particles are written in a manner similar
to (8.26).
Substituting (8.26) into (8.6) gives the vertical turbulent moisture ux near the
surface as
E
f

a
C
E
| v
h
(z
r
)| [ q
v
(z
0.v
) q
v
(z
r
)] (8.27)
(kg m
2
s
1
). This equation states that moisture is transferred upward when q
v
at
z
0.v
is greater than that at z
r
. The vertical eddy diffusion coefcient for moisture is
assumed to be the same as that for energy since C
E
is set to C
H
.
8.4.2 MoninObukhov similarity theory
In the presence of strong wind shear above the surface, bulk aerodynamic formulae
are not useful. A better method of parameterizing kinematic uxes near the surface
239
is with MoninObukhov similarity theory. Similarity theory is a method by which
variables are rst combined into dimensionless groups. Experiments are then con-
ducted to obtain values for each variable in the dimensionless group. The dimen-
sionless group, as a whole, is then tted, as a function of some parameter, with an
empirical equation. The experiment is repeated. Usually, equations obtained from
later experiments are similar to those fromthe rst experiment. Hence, this method
of obtaining an empirical equation for the dimensionless group is called similarity
theory, and the relationship between the empirical equation and the dimensionless
group is a similarity relationship. When similarity theory is applied to the sur-
face layer, it is usually called MoninObukhov similarity theory or surface-layer
similarity theory (Monin and Obukhov 1954; Stull 1988).
In this subsection, kinematic vertical turbulent uxes and eddy diffusion coef-
cients are derived from similarity theory. The derivation requires the discussion of
two similarity relationships and some parameters, discussed rst.
8.4.2.1 Dimensionless wind shear
One similarity relationship is that for the dimensionless wind shear,
m
k
=
z
u
| v
h
|
z
(8.28)
The right side of this equation, as a whole, is dimensionless. Individual factors, such
as the wind shear (| v
h
|z) and u
, are found from eld experiments. Wind shear

is measured directly, and u
is found from (8.8), the terms of which are measured.

The parameter
m
k is determined as a function of zL by substituting measure-
ments of | v
h
|z and u
into (8.28) for different values of zL and tting curves

to the resulting data. L is the MoninObukhov length (m), discussed shortly, and
zL is a dimensionless group. The von K arm an constant k is found by substitut-
ing measurements of | v
h
|z and u
into (8.28) under neutral conditions, when
m
= 1, then solving for k. Businger et al. (1971) derived
m
from eld data when
k = 0.35 as
m
=
1 +
m
z
L
z
L
> 0 stable
1
m
z
L
14
z
L
- 0 unstable
1
z
L
= 0 neutral
(8.29)
where
m
= 4.7 and
m
= 15.0. When k = 0.4, the values
m
= 6.0 and
m
=
19.3 should be used instead to obtain the same values of
m
as when k = 0.35
(Hogstrom 1988). Equation (8.29) was derived for the range, |zL| - 2, but has
been used successfully beyond the range under unstable conditions (San Jose et al.
1985). Other similarity relationships for
m
include those by Dyer (1974) and Dyer
and Bradley (1982), among others.
240
Integrating both sides of (8.28) between z
0.m
and z
r
, solving for u
, and noting
that | v
h
(z
0.m
)| = 0 give
u
=
k| v
h
(z
r
)|
z
r
z
0.m
m
dz
z
(8.30)
where
z
r
z
0.m
m
dz
z
=
ln
z
r
z
0.m
+

m
L
(z
r
z
0.m
)
z
L
> 0 stable
ln
1
m
z
r
L
14
1
1
m
z
r
L
14
+1
ln
1
m
z
0.m
L
14
1
1
m
z
0.m
L
14
+1
+2tan
1
1
m
z
r
L
14
2tan
1
1
m
z
0.m
L
14
z
L
- 0 unstable
ln
z
r
z
0.m
z
L
= 0 neutral
(8.31)
This integral exceeds zero and increases with increasing zL. Thus, for the same
wind speed, u
from (8.30) is larger in unstable air (zL - 0) than in stable air

(zL > 0). Increasing values of z
0.m
also increase u
by decreasing the integral

in (8.31).
8.4.2.2 MoninObukhov length
The MoninObukhov length (L) is a length scale (m) proportional to the height
above the surface at which buoyant production of turbulence rst dominates
mechanical (shear) production of turbulence. Mathematically,
L =
u
3
v
kg(w
v
)
s
=
u
2
v
kg
(8.32)
where
is a potential temperature scale (K), discussed shortly, and the second

expression is derived by substituting the similarity-theory approximation
(w
v
)
s
u
(8.33)
into the rst expression.
241
The parameter
is proportional to

v
(z
r
)

v
(z
0.h
), the vertical difference in
potential virtual temperature. The greater

v
at z
0.h
in comparison with its value
at z
r
, the more negative the change in

v
with increasing height, and the greater
the instability of the surface layer. In such cases, L is negative but has a small
magnitude, since it is inversely proportional to
. When L is negative with a small

magnitude, zL is negative with a large magnitude. Such values of zL correspond
to large instability due to buoyancy. Positive values of zL correspond to increasing
v
with altitude and stable stratication.
8.4.2.3 Dimensionless potential temperature gradient
An expression for the potential temperature scale,
, can be obtained from a

similarity relationship for the dimensionless potential temperature gradient,
h
k

z
v
z
(8.34)
where

v
z is the change in mean potential virtual temperature with height.
Businger et al. (1971) performed experiments to nd
h
for different stability
regimes when

p
was used instead of

v
and k = 0.35. The resulting parame-
terization was
h
=
Pr
t
+
h
z
L
z
L
> 0 stable
Pr
t
1
h
z
L
12
z
L
- 0 unstable
Pr
t
z
L
= 0 neutral
(8.35)
where
h
= 4.7,
h
= 9.0, and
Pr
t
=
K
m.zx
K
h.zz
(8.36)
is the turbulent Prandtl number, which approximates the ratio of the eddy diffusion
coefcient for momentumto that for energy. For k =0.35, Businger et al. estimated
Pr
t
0.74. Hogstrom (1988) noted that, when k = 0.4, Businger et al.s constants
should be modied to
h
= 7.8,
h
= 11.6, and Pr
t
0.95 to obtain the same
relationship as when k = 0.35. Integrating both sides of (8.34) between z
0.h
and z
r
and solving for
give
=
k[
v
(z
r
)

v
(z
0.h
)]
z
r
z
0.h
h
dz
z
(8.37)
242
where
z
r
z
0.h
h
dz
z
=
Pr
t
ln
z
r
z
0.h
+

h
L
(z
r
z
0.h
)
z
L
> 0 stable
Pr
t
ln
1
h
z
r
L
12
1
1
h
z
r
L
12
+1
ln
1
h
z
0.h
L
12
1
1
h
z
0.h
L
12
+1
z
L
- 0 unstable
Pr
t
ln
z
r
z
0.h
z
L
= 0 neutral
(8.38)
In sum, u
can be determined from (8.30) and (8.31) if L is known. If L is not

known, L, u
, and
must be determined simultaneously by solving (8.30), (8.32),

and (8.37).
8.4.2.4 Noniterative parameterization for momentum and potential
temperature scales
A noniterative method of determining u
, and L is with the parameterization

of Louis (1979). The rst step is to calculate the bulk Richardson number,
Ri
b
=
g[
v
(z
r
)

v
(z
0.h
)](z
r
z
0.m
)
2
v
(z
0.h
)[ u(z
r
)
2
+ v(z
r
)
2
](z
r
z
0.h
)
(8.39)
which quanties the ratio of buoyancy to mechanical shear. The second step is to
calculate u
and
as
u

k| v
h
(z
r
)|
ln(z
r
z
0.m
)
G
m

k
2
| v
h
(z
r
)|
v
(z
r
)

v
(z
0.h
)
Pr
t
ln
2
(z
r
z
0.m
)
G
h
(8.40)
where
G
m
= 1
9.4Ri
b
1 +
70k
2
(|Ri
b
| z
r
z
0.m
)
0.5
ln
2
(z
r
z
0.m
)
Ri
b
0
G
h
= 1
9.4Ri
b
1 +
50k
2
(|Ri
b
| z
r
z
0.m
)
0.5
ln
2
(z
r
z
0.m
)
Ri
b
0
G
m
. G
h
=
1
(1 +4.7Ri
b
)
2
Ri
b
> 0 (8.41)
243
The surface roughness length for momentum is used for momentum and energy
terms in (8.41) to maintain consistency. From values of u
and
, L can be deter-
mined diagnostically with (8.32).
Example 8.7
Given
z
0.m
= 0.01 m z
0.h
= 0.0001 m z
r
= 10 m
u(z
r
) = 10 ms
1
:(z
r
) = 5ms
1
k = 0.4
v
(z
r
) = 285 K

v
(z
0.h
) = 288 K Pr
t
= 0.95
calculate u
, and L.
SOLUTION
From (8.39)(8.41),
| v
h
(z
r
)| = 11.18ms
1
Ri
b
= 8.15 10
3
G
m
= 1.046
G
h
= 1.052 u
= 0.662 ms
1
= 0.188 K
L = 1.69 m
8.4.2.5 Gradient Richardson number
The bulk Richardson number is used for practical application in meteorological
modeling. It is derived from the gradient Richardson number,
Ri
g
=
g
v
z
u
z
2
+
v
z
2
(8.42)
The gradient and bulk Richardson numbers give the ratio of turbulence due to
buoyancy relative to that due to shear. When Ri
g
. Ri
b
- 0, the potential virtual
temperature decreases with increasing altitude, and the atmosphere is buoyantly
unstable and turbulent. When Ri
g
. Ri
b
are small and negative, wind shear is large in
comparison with buoyancy, and turbulence due to mechanical shear dominates tur-
bulence due to buoyancy. When Ri
g
. Ri
b
are large and negative, turbulence due to
free convection dominates turbulence due to forced convection. When Ri
g
. Ri
b
> 0,
the potential virtual temperature gradient exceeds zero, the atmosphere is buoy-
antly stable, and turbulence due to free convection does not occur. When Ri
g
. Ri
b
are small and positive, wind shear is large in comparison with buoyant stability,
and turbulence due to forced convection occurs. When Ri
g
. Ri
b
are large and pos-
itive, wind shear is low in comparison with buoyant stability, and turbulence due
to forced or free convection does not occur. Instead, air ow is laminar. Table 8.2
summarizes the ow regimes obtained from different Richardson numbers.
244
Table 8.2 Characteristics of vertical ow of air for different values of Ri
b
and Ri
g
Level of turbulence Level of turbulence
Ri
b
or Ri
g
Type of ow due to buoyancy due to shear
Large, negative Turbulent Large Small
Small, negative Turbulent Small Large
Small, positive Turbulent None (weakly stable) Large
Large, positive Laminar None (strongly stable) Small
The level of turbulence due to buoyancy is relative to that due to shear, and vice versa.
When Ri
b
and Ri
g
are large and positive and decrease to less than a critical
Richardson number (Ri
c
) of 0.25, laminar ow becomes turbulent and wind shear
increases. When Ri
b
and Ri
g
are small and positive and increase to above the
termination Richardson number (Ri
T
) of 1.0, turbulent ow becomes laminar.
8.4.2.6 Momentum uxes from similarity theory
Kinematic vertical turbulent uxes of momentum and eddy diffusion coefcients
for momentumcan be derived fromsimilarity theory. Substituting (8.16) and (8.17)
into (8.8) gives the relationship between friction wind speed and the coefcient of
drag as
u
= | v
h
(z
r
)|
C
D
(8.43)
Example 8.8
If C
D
= 0.001 and | v
h
(z
r
)| = 10 m s
1
, then u
= 0.32 m s
1
from (8.43). If
C
D
= 0.02 and | v
h
(z
r
)| = 10 m s
1
, then u
= 1.41 m s
1
. These are typical
values of u
over smooth and rough surfaces, respectively, when the surface

layer is well mixed.
Substituting the expression for C
D
from (8.43) into (8.16) and (8.17) gives the
kinematic vertical turbulent uxes of momentum from similarity theory as
(w
)
s
=
u
2
| v
h
(z
r
)|
u(z
r
) (8.44)
(w
)
s
=
u
2
| v
h
(z
r
)|
v(z
r
) (8.45)
Substituting (8.44), (8.45), and (8.19) into (8.18) and solving give the eddy diffu-
sion coefcients for momentum in the surface layer from similarity theory as
K
m.zx
= K
m.zy

u
2
| v
h
(z
r
)|
(z
r
z
0.m
) (8.46)
245
Further combining
| v
h
|
z

| v
h
(z
r
)|
z
r
z
0.m
(8.47)
and the dimensionless wind shear equation from (8.28) with (8.46) gives the dif-
fusion coefcient under stable, unstable, or neutral conditions as
K
m.zx
= K
m.zy

kzu
m
(8.48)
where kz =
e
is the mixing length of an eddy near the surface. The eddy mix-
ing length is the average distance an eddy travels before it exchanges momentum
with surrounding eddies. Near the surface, mixing is limited by the ground. Under
neutral conditions,
m
= 1, and (8.48) simplies to K
m.zx
= kzu
= K
m.zy
.
8.4.2.7 Energy and moisture uxes from similarity theory
The kinematic vertical turbulent uxes of sensible heat and moisture fromsimilarity
theory are
(w
v
)
s
= u
(8.49)
(w
v
)
s
= u
(8.50)
where
is the potential temperature scale (K) and q
is the water-vapor scale

(kg kg
1
). Like
, q
is found from a similarity relationship,
q
k
=
z
q
q
v
z
(8.51)
where
q

h
is a dimensionless specic humidity gradient. Substituting
h

q
into (8.51), integrating between z
0.v
and z
r
, and solving for q
give
q
=
k[ q
v
(z
r
) q
v
(z
0.v
)]
z
r
z
0.v
h
dz
z
(8.52)
This integral of
h
is found from (8.38), but with z
0.v
substituted for z
0.h
.
Substituting (8.49) and (8.50) into (8.4) and (8.6), respectively, gives the surface-
layer vertical turbulent sensible heat and moisture uxes as
H
f

a
c
p.d
u
E
f

a
u
(8.53)
246
(W m
2
and kg m
2
s
1
, respectively). Substituting (8.49) and (8.34) into (8.23)
gives the vertical eddy diffusion coefcient for energy from similarity theory as
K
h.zz
=
u
v
z)

kzu
h
(8.54)
(m
2
s
1
). The eddy diffusion coefcients for water vapor and other species are set
equal to those for energy.
Combining (8.48) and (8.38) with Pr
t
= K
m.zx
K
h.zz
from (8.36) gives the
similarity-theory expression,
Pr
t
=

h
m
(8.55)
Another parameterization for Pr
t
was developed by Tjernstrom (1993), who esti-
mated
Pr
t
(1 +4.47Ri
g
)
12
(8.56)
from turbulence observations in the boundary layer. In this equation, Ri
g
varies
from 0.01 to 10.
Example 8.9
Given the conditions in Example 8.7, nd K
h.zz
, K
m.zx
, K
m.zy
, and K
m.zx
K
h.zz
.
SOLUTION
Substituting u
, uz, :z, | v
h
(z
r
)|, and

v
z into (8.46) and (8.54) gives
K
h.zz
= 0.41 m
2
s
1
K
m.zx
= 0.39 m
2
s
1
K
m.zy
= 0.39 m
2
s
1
K
m.zx
K
h.zz
= 0.95
Thus, the Louis equations predict Pr
t
= K
m
K
h
consistently with the value of
Pr
t
used in (8.40).
8.4.2.8 Vertical proles of wind speed, potential virtual temperature,
and moisture
Similarity theory can be used to derive proles of vertical wind speed, potential
virtual temperature, and specic humidity in the surface layer. From (8.28) and
(8.34), the vertical gradients of wind speed and potential virtual temperature in the
247
surface layer are
| v
h
(z)|
z
=
u
kz
m
=
u
kz
[1 (1
m
)] (8.57)
v
z
=

kz
h
=

kz
[1 (1
h
)] (8.58)
respectively. Integrating the rst equation between z
0.m
and z and the second equa-
tion between z
0.h
and z gives wind speed and potential virtual temperature versus
altitude as
| v
h
(z)| =
u
ln
z
z
0.m
(8.59)
v
(z) =

v
(z
0.h
) +Pr
t
ln
z
z
0.h
(8.60)
respectively, where
m
=
z
z
0.m
(1
m
)
dz
z

h
=
z
z
0.h
(1
h
)
dz
z
(8.61)
are inuence functions for momentum and energy. Integrating (8.61) with values
of
m
and
h
from (8.29) and (8.35), respectively, gives
m
=
m
L
(z z
0.m
)
z
L
> 0 stable
ln
[1 +
m
(z)
2
][1 +
m
(z)
1
]
2
[1 +
m
(z
0.m
)
2
][1 +
m
(z
0.m
)
1
]
2
2tan
1
[
m
(z)]
1
+2tan
1
[
m
(z
0.m
)]
1
z
L
- 0 unstable
0
z
L
= 0 neutral
(8.62)
h
=
1
Pr
t
h
L
(z z
0.h
)
z
L
> 0 stable
2ln
1 +
h
(z)
1
1 +
h
(z
0.h
)
1
z
L
- 0 unstable
0
z
L
= 0 neutral
(8.63)
The inuence function for momentum accounts for the difference between a loga-
rithmic wind speed prole and an actual prole under stable and unstable condi-
tions. The inuence function for energy is analogous to that for momentum.
Under neutral conditions,
m
= 1, and (8.59) reduces to a standard logarithmic
wind prole for a neutrally stratied surface layer,
| v
h
(z)| =
u
k
ln
z
z
0.m
(8.64)
248
0
2
4
6
8
10
0 2 4 6 8 10 12
H
e
i
g
h
t

a
b
o
v
e

s
u
r
f
a
c
e

(
m
)
Wind speed (m s
1
)

0,
= 0.1 m
0,
= 1.0 m
z
m
z
m
Figure 8.3 Logarithmic wind proles in the
surface layer from (8.64) when u
= 1 m s
1
.
This equation states that the wind speed at z
0.m
is zero but increases logarithmically
with altitude. Figure 8.3 shows two examples of logarithmic wind proles.
The vertical prole of specic humidity in the surface layer is derived in the same
way as that for potential virtual temperature. The result is
q
v
(z) = q
v
(z
0.v
) +Pr
t
q
ln
z
z
0.v
(8.65)
where the energy inuence function from(8.63) is used, but with z
0.v
instead of z
0.h
.
In (8.60) and (8.65), values of the potential virtual temperature and specic humid-
ity at the ground surface are often substituted for values at the surface roughness
length.
In a canopy, such as in a eld of crops, an orchard, or a forest, the canopy
top affects wind speed, potential virtual temperature, and water vapor more than
does the ground. In the presence of a canopy, z
0.m
, z
0.h
, and z
0.v
are displaced a
vertical distance d
c
above the ground. This height is the displacement height, which
usually lies within 70 to 80 percent of the canopy height h
c
(Deardorff 1978; Pielke
1984; Kaimal and Finnigan 1994). When a displacement height exists, z
0.m
, z
0.h
,
and z
0.v
are dened as heights above the displacement height, and the mean wind
speed extrapolates to zero at the height d
c
+ z
0.m
. Wind speed, potential virtual
temperature, and specic humidity proles in a canopy are redened as
| v
h
(z)| =
u
ln
z d
c
z
0.m
z d
c
L
(8.66)
v
(z) =

v
(d
c
+ z
0.h
) +Pr
t
ln
z d
c
z
0.h
z d
c
L
(8.67)
q
v
(z) = q
v
(d
c
+ z
0.v
) +Pr
t
q
ln
z d
c
z
0.v
z d
c
L
(8.68)
respectively. Figure 8.4 shows the relationship among d
c
, h
c
, and z
0.m
and discusses
how to calculate d
c
. Table 8.1 gives values of h
c
and d
c
for some surfaces.
249
ln(d
c
+ z
0,m
)
0
0
|v
h
|
ln d
c
ln h
c
Top of canopy
l
n

z
Figure 8.4 Relationship among d
c
, h
c
,
and z
0.m
. The displacement height is
found by plotting wind speed over a
canopy versus ln z. The plotted wind
speed is extrapolated to zero to obtain
d
c
+ z
0.m
. Different values of d
c
are sub-
stituted into d
c
+ z
0.m
to estimate z
0.m
.
Both d
c
and z
0.m
are substituted into
(8.66) until the predicted curve of wind
speed with height matches the logarith-
mic curve shown in the diagram. This
method works best when winds aloft are
strong.
Another common way to estimate the vertical prole of wind speed in the bound-
ary layer is with a power-law prole,
| v
h
(z)| = | v
h
(z
r
)|
z
z
r
(8.69)
where is typically set to a constant, such as 17 (e.g., Elliott et al. 1986; Arya
1988). Both the power law and logarithmic prole can replicate observed wind
proles in idealized cases. However, real wind proles are often far from ideal.
For some purposes, it is thus useful to t observed wind proles with analytical
expressions (e.g., Archer and Jacobson 2003).
8.5 EDDY DIFFUSION ABOVE THE SURFACE LAYER
Similarity theory expressions for eddy diffusion coefcients are used for the surface
layer. Above the surface layer, turbulent transport of momentum and energy can
be parameterized with a hybrid local/nonlocal closure scheme or with a scheme
that solves the turbulent kinetic energy (TKE) equation to high order. Below, both
hybrid and TKE schemes are discussed.
8.5.1 Hybrid scheme
In a hybrid model, vertical turbulent transport is treated differently for stable and
weakly unstable conditions versus for strongly unstable conditions. For stable and
250
8.5 Eddy diffusion above the surface layer
weakly unstable conditions, turbulent transport is simulated with K-theory using
an eddy diffusion coefcient for momentum that depends on mechanical shear and
buoyancy. One such diffusion coefcient is
K
m.zx
K
m.zy

2
e
L u
Lz
2
+
L v
Lz
2
Ri
c
Ri
b
Ri
c
(8.70)
(m
2
s
1
) (e.g., Blackadar 1976; Stull 1988), where z is in meters and
e
=
kz
1 +kz
m
(8.71)
is an expression for the mixing length that simplies to kz near the surface
(small z) and to the free-atmospheric mixing length,
m
= 70200 m, above the
surface layer (large z). Near the surface, turbulence and mixing length are limited
by the ground. In the free atmosphere, turbulence is limited by a maximum mixing
length. Equation (8.70) is applicable when Ri
b
- Ri
c
0.25. When Ri
b
Ri
c
, the
atmosphere is strongly stable, (8.70) predicts no turbulence (a negative value of
K
m.zx
), and the diffusion coefcient is set to a minimum positive value.
When Ri
c
> Ri
b
> 0, turbulence is due to mechanical shear, and the atmosphere
is weakly stable. When Ri
b
is small and negative, turbulence is due to shear and
weak buoyancy, and the atmosphere is weakly unstable. Under both conditions,
(8.70) can be used.
In a jet, wind speeds increase and then decrease with altitude, minimizing wind
shear in the jets peak. In such a case, and in the absence of buoyancy, (8.70) predicts
a small K
m.zx
. Some observations of turbulence near a jets maximumspeed indicate
that turbulence peaks near the maximum(Lenschowet al. 1988; Tjernstrom1993),
while other observations indicate that turbulence is lowest near the jet maximum
(e.g., Mahrt et al. 1979; Lenschow et al. 1988; Tjernstrom 1993). In cases when
the turbulence peaks near the jet maximum, (8.70) may not properly predict K
m.zx
(Shir and Bornstein 1976).
The eddy diffusion coefcient for energy is related to that for momentum by
K
m.zx
K
m.zy
Pr
t
K
h.zz
. Eddy diffusion coefcients for trace gases or particles
are usually set equal to those for energy.
When Ri
b
is large and negative, the atmosphere is strongly unstable, and free
convection occurs. Since (8.70) captures effects of small eddies but not large eddies
that arise during free convection (Stull 1988), it is not used to simulate free con-
vection. Instead, vertical turbulent transport under strong, unstable conditions is
often simulated with a free convective plume scheme (Blackadar 1978; Zhang
and Anthes 1982; Lu 1994). A free convective plume scheme assumes buoyant
plumes from the surface rise and mix with air in every level of the boundary layer,
exchanging momentum, energy, moisture, and gases. The result of mixing is to
distribute these parameters evenly throughout the boundary layer in a short time.
A plume scheme differs from a K-theory parameterization in that the former mixes
all layers simultaneously whereas the latter mixes material between adjacent layers.
251
A turbulence scheme that mixes material among all layers simultaneously is called
a nonlocal closure scheme. One that mixes material between adjacent grid layers
is called a local closure scheme.
8.5.2 TKE schemes
Several local-closure turbulence parameterizations have been developed to predict
diffusion coefcients as a function of turbulent kinetic energy (TKE) for use in
K-theory equation models. In such cases, TKE itself is modeled with a time-
dependent equation. Two major types of TKE models have been developed, E-
e
and E-
d
models, where E is TKE,
e
is the mixing length, and
d
is the dissipation
rate of turbulent kinetic energy. The main difference between the two types is that
E-
e
models solve prognostic equations for E and
e
, whereas E-
d
models solve
prognostic equations for E and
d
. Both types of models are local closure models,
generally of order 2, 2.5, 3, or 4 (e.g., Mellor and Yamada 1974, 1982; Andre et al.
1978; Enger 1986; Briere 1987; Andr en 1990; Apsley and Castro 1997; Abdella
and McFarlane 1997; Liu and Leung 2001; Cheng et al. 2002; Freedman and
Jacobson 2002, 2003).
8.5.2.1 E-
e
schemes
In this subsection, the order 2.5 E-
e
model of Mellor and Yamada (1982) is
discussed. In this scheme, TKE is modeled prognostically with
E
t

z
s
q
2E
E
z
= P
s
+ P
b
d
(8.72)
where E = (u
2
+v
2
+w
2
)2 is TKE per unit mass (m
2
s
2
), s
q
is a constant set to
0.2,
e
is the eddy mixing length (m), s
q
2E is a diffusion coefcient for turbulent

transport of the gradient of TKE (m
2
s
1
), P
s
and P
b
are the production rates of
TKE due to shear and buoyancy, respectively (m
2
s
3
), and
d
is the dissipation
rate of TKE to heat due to molecular viscosity (m
2
s
2
s
1
= m
2
s
3
).
The eddy mixing length is determined from the prognostic equation for E
e
,
(2E
e
)
t

z
s
l
2E
(2E
e
)
z
=
e
e
1
(P
s
+ P
b
)
e
1 +e
2
e
kz
(8.73)
where s
l
is a constant set to 0.2, s
l
2E is a diffusion coefcient for turbulent

transport of the gradient of E
e
(m
2
s
1
), e
1
= 1.8 and e
2
= 1.33 are constants,
k = 0.4 is the von K arm an constant, and z is the height above the surface (m).
The mixing length is extracted from this equation with
e
= E
e
E, where E in
252
8.5 Eddy diffusion above the surface layer
the denominator is determined from (8.72). As a simplication, the mixing length
can also be determined diagnostically from (8.71).
In (8.72) and (8.73), the turbulent production rates of shear and buoyancy are
P
s
= K
m
u
z
2
+
v
z
(8.74)
P
b
=
g
v
K
h
v
z
(8.75)
respectively, where K
m
and K
h
are diffusion coefcients for momentum and energy
(m
2
s
1
) from the previous time step of calculation. The dissipation rate of TKE is
d
=
(2E)
32
B
1
e
(8.76)
where B
1
is a constant set to 16.6. Once E and
e
have been calculated, new
diffusion coefcients are determined with
K
m
= S
M
2E (8.77)
K
h
= S
H
2E (8.78)
where S
M
and S
H
are calculated from the following equations:
S
M
= 2E
0.74P
3
+0.6992P
1
P
1
P
4
P
3
P
2
(8.79)
S
H
= 2E
0.74P
4
+0.6992P
2
P
1
P
4
P
3
P
2
(8.80)
P
1
= 2E 30.5916G
H
(8.81)
P
2
= 4.0848G
M
(8.82)
P
3
= 16.284G
H
(8.83)
P
4
= 6.1272G
H
5.0784G
M
2E (8.84)
G
M
=

2
e
2E
u
z
2
+
v
z
(8.85)
G
H
=
2
e
2E
g
v
z
(8.86)
The last two terms are related to the gradient Richardson number by
Ri
g
=
G
H
G
M
(8.87)
253
Diffusion coefcients are used not only to recalculate production rates of shear and
buoyancy for the next time step but also to determine kinematic vertical turbulent
uxes of momentum and sensible heat.
8.5.2.2 E-
d
schemes
The main difference between an E-
e
and an E-
d
scheme is that, in the latter, a
prognostic equation is solved for the dissipation rate of TKE instead of for the
mixing length. Prognostic equations for TKE are solved in both cases. One form
of the prognostic equation for the dissipation rate is
d
t

z
K
m
d
z
= c
1
d
E
(P
s
+ P
b
) c
2
2
d
E
(8.88)
where
K
m
= c
j
E
2
d
(8.89)
is an expression for the eddy diffusion coefcient for momentum, and
, c
1
,
c
2
, and c
j
are constants determined either from experiment or from enforcing
consistency with MoninObukhov similarity theory (e.g., Freedman and Jacobson
2002, 2003). The mixing length for use in the TKE equation (8.72) is calculated
diagnostically from the most recent value of the TKE and the dissipation rate
with
e
= c
34
j
E
32
d
(8.90)
8.6 GROUND SURFACE TEMPERATURE AND
SOIL MOISTURE
Ground temperature and soil moisture (liquid-water content) affect the uxes of
energy and moisture, respectively, from the ground to the boundary layer. The
uxes, in turn, affect mixing height, wind speed, and pollutant concentration. For
example, low soil moisture increases ground temperature, increasing thermal tur-
bulence, increasing mixing height, increasing wind speed, and decreasing primary
pollutant concentration (Jacobson 1999a). The increase in near-surface wind speed
is due to the increased turbulent transport of momentumfromaloft, where the wind
is fast, to the surface, where it is generally slower. In this section, factors that affect
surface temperature and moisture content are discussed, and equations describing
soil energy and liquid-water transport are described.
8.6.1 Factors affecting soil temperature
Soil temperature is affected by several parameters, discussed briey below.
254
8.6 Ground surface temperature and soil moisture
8.6.1.1 Specic heat
Dry soil contains solid soil and air (voids). Since the specic heat of dry air is
lower than that of solid soil (Table 2.2), the average specic heat of soil plus air
is less than that of soil alone. When liquid water is added to soil, it replaces air.
Since the specic heat of liquid water is much larger than that of air, the specic heat
of a soilwaterair mixture is greater than that of a soilair mixture of the same
volume. Thus, a wet, sandy soil heats up less during the day and cools down less
during the night than does a dry, sandy soil when only specic heat is considered.
8.6.1.2 Soil moisture
Soil moisture also affects the rate of evaporation. The lower the liquid-water con-
tent of soil, the lower the rate of evaporation of water to the air (or greater the
rate of condensation of water vapor to the ground), the lesser the latent heat ux
to the air, and the lesser the cooling (greater the warming) of the ground due to
absorption (release) of latent heat by liquid water when it evaporates (condenses).
8.6.1.3 Conduction
Conduction between surface soil and molecules below the surface affects soil tem-
perature. During the night, the soil surface cools radiatively, creating a temperature
gradient in the top soil layers, forcing energy stored below the surface to conduct
upward to replenish the lost energy. The greater the thermal conductivity of the
soil, the faster the energy transfer occurs. Table 2.2 shows that clay is more con-
ductive than sand. At night, the replenishment of energy to a clay surface from the
subsoil is faster than for a sandy surface. During the day, conduction of absorbed
radiation from the surface to the subsoil is faster for clay than for sand.
8.6.1.4 Additional factors
Additional factors affect soil moisture and ground temperatures. Solar radiation
heats the ground during the day. Infrared emission cools the ground during the day
and the night. Vegetation cover reduces the solar radiation reaching the ground.
Transpiration removes liquid water fromdeep soil covered by vegetation. The water
is drawn fromsoil through roots and xylem(water-conducting tissue within plants)
to the leaves. Water vapor escapes through plant leaf stomata (pores). Finally, a
major source of liquid water in soil is precipitation.
Below, equations for calculating bare-soil temperature and moisture are given.
Subsequently, equations for calculating temperature and moisture of vegetated soil
and other surfaces are described.
8.6.2 Ground temperature and moisture over bare soil
Soil models predict soil surface and subsurface temperature and liquid water con-
tent by dividing the soil near the surface into multiple layers. In this subsection,
255
equations for temperature and moisture changes among multiple soil layers are
described for when the soil surface is bare.
At the soil surface, conductive, radiative, sensible, and evaporative energy uxes
affect soil temperature. Below the surface, conduction is the most important factor
affecting temperature. Temperature changes due to conduction in a homogeneous
soil below the surface are estimated with the heat conduction equation,
T
s
t
=
1
g
c
G
s
T
s
z
(8.91)
where T
s
is the soil temperature,
s
is the thermal conductivity of the soilwaterair
mixture (J m
1
s
1
K
1
),
g
is the density of the mixture (kg m
3
), c
G
is the specic
heat of the mixture (J kg
1
K
1
), and
s
T
s
z is the conductive heat ux through
the soilwaterair mixture (J m
2
s
1
). The thermal conductivity (J m
1
s
1
K
1
)
of a soilwaterair mixture may be approximated with
s
= max
418e
log
10
|
p|2.7
. 0.172
(8.92)
(Al Nakshabandi and Konhke 1965; McCumber and Pielke 1981), where
p
is the
moisture potential (cm) or soil water tension and max indicates the larger of
the two values. The moisture potential is the potential energy required to extract
water from capillary and adhesive forces in the soil. Clapp and Hornberger (1978)
parameterized the moisture potential as
p
=
p.s
w
g.s
w
g
b
(8.93)
where
p.s
is the moisture potential when the soil is saturated with liquid water
(cm), w
g
is the volumetric water content of the soil (soil moisture) in cubic meters
of liquid water per cubic meter of soilwaterair mixture, w
g.s
is the maximum
volumetric water content that a given soil type can hold (m
3
m
3
), and b is a
coefcient required to t (8.93) to data. Values of
p.s
, w
g.s
, and b are given in
Table 8.3 for different soil types. An alternative to the Clapp and Hornberger
moisture potential equation is one developed by van Genuchten (1980) that has
been widely used in soil-physics studies (Cuenca et al. 1996).
The product of mass density and specic heat of a soilwaterair mixture is
g
c
G
= (1 w
g.s
)
s
c
S
+w
g
w
c
W
(8.94)
where
s
is the density of solid soil (kg m
3
), c
S
is the specic heat of solid soil
(J kg
1
K
1
),
w
is the density of liquid water (1000 kg m
3
), c
W
is the specic
heat of liquid water (J kg
1
K
1
), 1 w
g.s
is the volumetric content of solid soil,
w
g
is the volumetric content of water in the soil, and w
g.s
w
g
is the volumetric air
content in the soilwaterair mixture. In the equation, the product of mass density
and specic heat of air is neglected, since the mass density of air is much smaller
than that of soil or water. Values of
s
c
S
are given in Table 8.3 for different soil
types.
256
Table 8.3 Soil parameters for 11 soil types
w
g.s
w
fc
w
wilt

p.s
K
g.s

s
c
S
Soil type
b (m
3
m
3
) (m
3
m
3
) (m
3
m
3
) (cm) (m s
1
) (J m
3
K
1
)
Sand 4.05 0.395 0.135 0.068 12.1 1.76(4) 1.47(6)
Loamy sand 4.38 0.410 0.150 0.075 9.0 1.56(4) 1.41(6)
Sandy loam 4.90 0.435 0.195 0.114 21.8 3.41(5) 1.34(6)
Silt loam 5.30 0.485 0.255 0.179 78.6 7.20(6) 1.27(6)
Loam 5.39 0.451 0.240 0.155 47.8 7.00(6) 1.21(6)
Sandy clay loam 7.12 0.420 0.255 0.175 29.9 6.30(6) 1.18(6)
Silty clay loam 7.75 0.477 0.322 0.218 35.6 1.70(6) 1.32(6)
Clay loam 8.52 0.476 0.325 0.250 63.0 2.50(6) 1.23(6)
Sandy clay 10.40 0.426 0.310 0.219 15.3 2.20(6) 1.18(6)
Silty clay 10.40 0.492 0.370 0.283 49.0 1.00(6) 1.15(6)
Clay 11.40 0.482 0.367 0.286 40.5 1.30(6) 1.09(6)
Adapted from Clapp and Hornberger (1978), Pielke (1984), Noilhan and Planton (1989), and
Mahfouf and Noilan (1996). 1.00(6) means 1.00 10
6
.
The time rate of change of volumetric water content of soil below the surface
can be approximated with the water transport equation,
w
g
t
=

z
K
g
p
z
+1
=

z
D
g
w
g
z
+ K
g
(8.95)
(McCumber and Pielke 1981) where K
g
is the coefcient of permeability of liquid
water through soil (hydraulic conductivity), D
g
is the diffusion coefcient for water
in soil, and K
g
(
p
+ z)z is the kinematic ux of liquid water through the soil
(m [m
3
m
3
] s
1
). Liquid water feeds back to soil temperature through the thermal
conductivity and specic heat terms in (8.91).
The hydraulic conductivity (ms
1
), which accounts for gravity drainage through
a viscous soil, is affected by water viscosity and the shapes and sizes of voids
between soil particles. Clapp and Hornberger (1978) parameterized it as
K
g
= K
g.s
w
g
w
g.s
2b+3
(8.96)
where K
g.s
is the hydraulic conductivity at saturation (m s
1
), and values of b
and w
g.s
are shown in Table 8.3. An alternative parameterization is given by van
Genuchten (1980). The diffusion coefcient for water in soil (m
2
s
1
) is
D
g
= K
g
p
w
g
=
bK
g.s
p.s
w
g
w
g
w
g.s
b+3
=
bK
g.s
p.s
w
g.s
w
g
w
g.s
b+2
(8.97)
where the second expression was obtained by substituting (8.93) and (8.96) into
the rst.
257
At the soil surface, (8.91) and (8.95) can be modied to
T
s
t
=
1
g
c
G
s
T
s
z
+ F
n.g
H
f
L
e
E
f
(8.98)
w
g
t
=

z
D
g
w
g
z
+ K
g
+
E
f
P
g
(8.99)
respectively, where
s
T
s
z (W m
2
) is the conductive heat ux between the soil
surface and the layer of soil just below the surface, F
n.g
(W m
2
) is the net down-
ward minus upward solar plus infrared radiative ux at the surface (positive F
n.g
is down), H
f
(W m
2
) is the vertical turbulent sensible-heat ux at the soil surface
(positive is up), E
f
(kg m
2
s
1
) is the evaporation rate at the surface (positive is
up), L
e
(J kg
1
) is the latent heat of evaporation, and P
g
(kg m
2
s
1
) is the net
ux of liquid water reaching the soil surface (precipitation minus runoff) (posi-
tive is down). The product L
e
E
f
is the net latent heat ux between the soil sur-
face and the atmosphere. F
n.g
is determined from a radiative-transfer calculation
(Chapter 9).
H
f
and E
f
can be estimated from similarity theory with (8.53) or from bulk
aerodynamic equations with (8.22) and (8.27), respectively. H
f
depends on the
potential virtual temperature of air at height z
0.h
, and E
f
depends on the specic
humidity of air at height z
0.v
. Two methods of calculating the specic humidity at
height z
0.v
are
q
v
(z
0.v
) =
g
q
v.s
(T
g
) (8.100)
q
v
(z
0.v
) =
g
q
v.s
(T
g
) +(1
g
) q
v
(z
r
) (8.101)
where q
v.s
(T
g
) is the saturation specic humidity at the ground temperature, and
g
and
g
are wetness functions. The function
g
is like a relative humidity adjacent
to the water in soil pores. Philip (1957) estimated
g
exp[
p
gR
v
T
g
], but Wetzel
and Chang (1987), Avissar and Mahrer (1988), Kondo et al. (1990), Lee and Pielke
(1992), and Mihailovic et al. (1995) pointed out that this equation is incorrect.
These papers and Mahfouf and Noilhan (1991) give alternative formulations for
g
. An expression for
g
is
g
=
1
4
1 cos
w
g
w
fc
2
w
g
- w
fc
and q
v
(z
r
) q
v
(z
0.v
)
1 w
g
w
fc
or q
v
(z
r
) > q
v
(z
0.v
)
(8.102)
(Lee and Pielke 1992), who obtained the t from data of Kondo et al. (1990),
where w
g
(m
3
m
3
) is the liquid water content (LWC) of the top soil layer and w
fc
(m
3
m
3
) is the soil LWC at eld capacity (Table 8.3), which is the LWC after all
macropores of soil have emptied due to gravitational drainage and liquid water
remains only in soil micropores. When condensation occurs, and for water, ice,
and snow surfaces,
g
= 1.
Together, (8.91), (8.95), (8.98), and (8.99) can be solved numerically among
several soil layers (e.g., McCumber and Pielke, 1981). The solution is generally
258
obtained explicitly for below-surface layers assuming (8.99) applies to the middle
of the top soil layer (replacing (8.95) in that layer). At the soil surface, ground
temperature is then diagnosed from the surface energy balance equation,
s.1
T
s
z
+ F
n.g
H
f
L
e
E
f
= 0 (8.103)
which is obtained by assuming (8.98) is in steady state. In this equation,
s.1
(W m
1
K
1
) is the thermal conductivity of a soilwaterair mixture in the top
model-layer of soil. The soil-top temperature from the equation is used as an upper
boundary term during the next time step in the calculation of subsurface tempera-
tures from (8.91).
An alternative to a multilayer soil model is a two-compartment soil model (e.g.,
Noilhan and Planton 1989; Mahfouf and Noilhan 1996). This model is a gener-
alization of the force-restore method, which involves forcing the temperature and
liquid-water content in the top centimeter of soil over a short time and restor-
ing these variables with deep-soil values over a longer time (Bhumralkar 1975;
Blackadar 1976; Deardorff 1977).
8.6.3 Temperature and moisture in vegetated soil
When soil is covered with vegetation, energy and moisture from the soil must pass
through the vegetation before reaching the free atmosphere. Conversely, energy and
moisture fromthe atmosphere must pass through the vegetation before reaching the
ground. In this section, energy and moisture transfer through and storage within
vegetation are discussed. The numerical treatment is described in more detail in
Jacobson (2001a).
Vegetation, as dened here, includes trees, shrubs, grass, and plants. A canopy
is the vegetation cover provided by a group of plants or trees. Foliage is a cluster
of leaves within a canopy.
Ground temperature in a model grid cell that contains bare soil and soil covered
by vegetation can be calculated with the surface energy balance equation from
(8.103), modied with the following uxes:
F
n.g
= f
s

F
s
+

F
i

s
B

T
4
g
(8.104)
H
f
= f
s

a
c
p.d
R
a
p
(z
r
)
T
g
P
g
f
v

a
c
p.d
R
f
T
af
P
f
T
g
P
g
(8.105)
L
e

E
f
= f
s
L
e

a
R
a
g
[ q
v
(z
r
) q
v.s
(

T
g
)] f
v
L
e

a
R
f
g
[q
af
q
v.s
(

T
g
)]
(8.106)
259
where
F
n.g
Grid-cell averaged net downward minus upward solar plus infrared radia-
tive ux absorbed by bare soil (positive is down) (W m
2
)
H
f
Grid-cell averaged vertical turbulent sensible-heat ux. It is calculated as
the ux between bare soil and open air plus that between vegetated soil
and canopy air (positive is up) (W m
2
)
E
f
Grid-cell averaged water vapor ux between bare soil and open air plus
that between vegetated soil and canopy air (positive is up) (kg m
2
s
1
)
L
e
Latent heat of evaporation (J kg
1
)
f
s
, f
v
Fractions of bare andvegetatedsoil, respectively, ina gridcell ( f
s
+ f
v
= 1)
F
s
Grid-cell averaged net downward minus upward solar irradiance at the
top of the canopy plus that over bare soil (positive is downward) (W m
2
)
f
s

F
s
Net downward minus upward solar irradiance over bare soil in a grid cell
(positive is downward) (W m
2
)
F
i
Grid-cell averaged downward (not net downward minus upward)
thermal-IR irradiance absorbed by the ground (positive is downward)
(W m
2
)
T
g
Grid-cell averaged soil surface temperature (the surface value of T
s
) (K)
B
StefanBoltzmann constant (5.67051 10
8
W m
2
K
4
)
s
Thermal-IR emissivity of soil ()
s

T
4
g
Grid-cell averaged upward thermal-IR irradiance emitted by the ground
(W m
2
)

a
Grid-cell averaged air density immediately above the surface (kg m
3
)
c
p.d
Specic heat of dry air at constant pressure (J kg
1
K
1
)
R
a
Aerodynamic resistance (s m
1
) between bare soil outside of a canopy and
a reference height z
r
R
f
Aerodynamic resistance between soil and overlying vegetation (s m
1
)
p
(z
r
) Grid-cell averaged potential temperature at the reference height (K)
P
g
[ p
a.g
1000 hPa]
converts temperature at the ground to potential temper-

ature
p
a.g
Air pressure at the ground (hPa)
R
c
p.d
= 0.286
P
f
[ p
a.f
1000 hPa]
converts temperature in the foliage to potential tempera-

ture
T
af
Temperature of air in foliage (K)
g
Wetness function from (8.92)
q
v
(z
r
) Grid-cell averaged water vapor specic humidity at the reference height
(kg kg
1
)
q
v.s
(T
g
) Saturation specic humidity at the ground temperature (kg kg
1
)
q
af
Specic humidity of water vapor in foliage air (kg kg
1
)
z
r
Reference height (10 m)
An expression for the aerodynamic resistance (s m
1
) over bare soil is given in
(20.12). An expression for the aerodynamic resistance in the foliage (s m
1
) is
R
f
=
1
c
f
u
af
(8.107)
260
where u
af
0.83u
(m s
1
) is the wind speed in the foliage and
c
f
= 0.01
0.003
u
af
(8.108)
(u
af
in m s
1
) is the dimensionless heat-transfer coefcient (e.g., Deardorff 1978;
Pielke 1984). Expressions for temperature and specic humidity of air in the foliage
are
T
af
= 0.3

T
a
(z
r
) +0.6T
f
+0.1

T
g
(8.109)
q
af
= 0.3 q
v
(z
r
) +0.6q
f
+0.1 q
g
(8.110)
(Deardorff 1978), where
T
a
(z
r
) Grid-cell averaged temperature at the reference height (K)
q
v
(z
r
) Grid-cell averaged water vapor specic humidity at reference height
(kg kg
1
)
T
f
Temperature in leaf stomata (K)
q
f
Specic humidity in leaf stomata (kg kg
1
)
T
g
Grid-cell averaged ground temperature (K)
q
g
Grid-cell averaged specic humidity in soil macropores (kg kg
1
)
The water vapor specic humidity in leaf stomata is
q
f
= min[ q
v.s
(T
f
) +(1 )q
af
. q
v.s
(T
f
)]. (8.111)
where
=
R
f
R
f
+ R
st
W
c
W
c.max
23
q
af
- q
v.s
(T
f
)
0 q
af
q
v.s
(T
f
)
(8.112)
(Deardorff 1978). In this equation, W
c
(m
3
m
2
or m) is the depth of liquid water
on an individual leaf surface (m) multiplied by the one-sided leaf area index, L
T
(square meters of leaf surfaces per square meter of underlying ground), W
c.max
(m) is the maximum possible value of W
c
(approximated as 0.0002L
T
(m) from
Dickinson 1984), and
R
st
=
R
min
F
c
1 +
200
F
s
+0.1
400
(T
f
273.15) (313.15 T
f
)
(8.113)
is the leaf stomata resistance (s m
1
), which is innite when T
f
273.15 K and
T
f
313.15 K, and large at night, when

F
s
= 0 (Baldocchi et al. 1987). In this
equation, R
min
is the minimum bulk canopy stomata resistance (Appendix Table
B.11), and
F
c
= max
min
w
g.avg
w
wilt
w
cr
w
wilt
. 1.0
. 1.0 10
12
(8.114)
is a factor accounting for reduction in transpiration due to drying up of the soil
toward the wilting point, w
wilt
(m
3
m
3
) (e.g., Noilhan and Planton 1989), which
261
is the liquid water content of soil when a plant is permanently wilted. Wilting
points of different soils are given in Table 8.3. In this equation, w
g.avg
(m
3
m
3
) is
the average liquid water content in the root layers of soil, and w
cr
(m
3
m
3
) is a
critical liquid-water content taken as 0.75w
g.s
(Thompson et al. 1981), where w
g.s
(m
3
m
3
) is the soil moisture content at saturation.
The foliage temperature T
f
is found by solving iteratively a foliage energy balance
equation that considers a net solar ux at the top of the foliage, a net thermal-IR
ux at the top of the foliage and at ground level, and sensible and latent heat uxes
at the top of the foliage and at ground level (e.g., Deardorff 1978; McCumber
1980; Pielke 1984; Jacobson 2001a). One such equation is
f
v

F
s
+
v

F
i
+

v
v
+
s
B

T
4
g

v
+2
s
v
+
s
B
T
4
f
= H
v
+ L
e
E
d
+ L
e
E
t
(8.115)
where
v
is the thermal-IR emissivity of vegetation,

F
i
(W m
2
) is the down-
ward thermal-IR irradiance at the top of the canopy, H
v
(W m
2
) is the sensible
heat ux between air in the foliage and leaves, E
d
(kg m
2
s
1
) is the turbulent
moisture ux due to direct evaporation from/condensation to leaves in the foliage,
and E
t
(kg m
2
s
1
) is the turbulent moisture ux due to transpiration from leaf
stomata.
The sensible heat ux, direct evaporation/condensation, and transpiration terms
in (8.114) are
H
v
= 1.1L
T

a
c
p.d
R
f
P
f
[T
af
T
f
] (8.116)
E
d
= L
T

a
d
R
f
[q
af
q
v.s
(T
f
)] (8.117)
E
t
= L
T

a
(1
d
)
R
f
+ R
st
[q
af
q
v.s
(T
f
)] (8.118)
respectively, where the factor 1.1 in (8.116) accounts for the effects of stalks,
stems, twigs, and limbs that exchange energy but do not transpire, L
T
is limited to
no greater than 7f
v
, and
d
=
W
c
W
c.max
23
q
af
- q
v.s
(T
f
)
1 q
af
q
v.s
(T
f
)
(8.119)
accounts for the decrease in the evaporation rate when the leaves contain little
water (Monteith and Szeicz 1962; Deardorff 1978). During condensation, direct
moisture ux to leaf surfaces via (8.117) occurs at the potential rate (
d
=1). When
dewcondenses on leaves, 1
d
=0 and no transpiration occurs by (8.118). When
direct evaporation occurs, the transpiration rate is limited by the fraction of leaf
surfaces not covered by liquid water. The saturation specic humidities in (8.117)
and (8.118) are obtained at the temperature of the foliage.
262
The iterative solution to T
f
is found by substituting (8.116)(8.118) into (8.114),
then linearizing T
4
f
and q
v.s
(T
f
) with
T
4
f.t.n+1
= T
4
f.t.n
+4T
3
f.t.n
(T
f.t.n+1
T
f.t.n
) (8.120)
q
v.s
(T
f.t.n+1
) = q
v.s
(T
f.t.n
) +
dq
v.s
(T
f.t.n
)
dT
(T
f.t.n+1
T
f.t.n
) (8.121)
respectively (Deardorff 1978), where the subscript t indicates a value at the end of
the time step and the subscript n indicates iteration number. Applying these steps
gives the foliage temperature as
T
f.t.n+1
=
f
v
F
s
+
v

F
i
+

v
v
+
s
B

T
4
g.th
+3
v
+2
s
v
+
s
B
T
4
f.t.n
+1.1L
T

a
c
p.d
R
f
P
f
T
af
+ L
e
L
T

a
d
R
f
+

a
t
R
f
+ R
st
q
af
q
v.s
(T
f.t.n
)
dq
v.s
(T
f.t.n
)
dT
T
f.t.n
v
+2
s
v
+
s
B
T
3
f.t.n
+1.1L
T

a
c
p.d
R
f
P
f
+L
e
L
T

a
d
R
f
+

a
(1
d
)
R
f
+ R
st
dq
v.s
(T
f.t.n
)
dT
(8.122)
(Jacobson 2001a), where the subscript t h indicates a value at the beginning of a
time step. This equation needs to be iterated only four times to achieve signicant
convergence.
A prognostic equation for water on leaf surfaces is
W
c
t
=

P
r
E
d
w
R (8.123)
where

P
r
(ms
1
) is the grid-cell averaged precipitation rate,
w
(kg m
3
) is the den-
sity of liquid water, and R (m s
1
) is the runoff rate from leaf surfaces. Combining
(8.117), (8.119), and (8.123) and nite-differencing the result over time step h (s)
give
W
c.t
= W
c.th
+
P
r
+
L
T
R
f

a
W
c.th
W
c.max
2
3
[q
af
q
v.s
(T
f
)]
q
af
- q
v.s
(T
f
)
h
P
r
+
L
T
R
f

a
w
[q
af
q
v.s
(T
f
)]
q
af
q
v.s
(T
f
)
(8.124)
The solution is limited by W
c
= min(max[W
c
. 0]. W
c.max
), which accounts for the
fact that any liquid water over depth W
c.max
is runoff (thus R does not appear in
the equation). Each solution from(8.124) is substituted into (8.119), which is itself
263
substituted into (8.117) to give the direct evaporation rate and into (8.118) to give
the transpiration rate.
The grid-cell averaged soil-top temperature, accounting for bare and vegetated
soil, is now calculated by substituting (8.104)(8.106) into the energy balance
equation (8.103), and linearizing. The result is
T
g.t.n
=

T
g.t.n1
+
f
s

a
c
p.d
R
a
p
(z
r
)
T
g.t.n1
P
g
f
v

a
c
p.d
R
f
T
af.t
P
f
T
g.t.n1
P
g
f
s

a
L
e
R
a
g
[ q
v
(z
r
) q
v.s
(

T
g.t.n1
)]
f
v

a
L
e
R
f
g
[q
af
q
v.s
(

T
g.t.n1
)]
f
s

F
s
+

F
i

B
s

T
4
g.t.n1
s.1
D
1
(T
1.t

T
g.t.n1
)
f
s

a
c
p.d
R
a
P
g
+ f
v

a
c
p.d
R
f
P
g
+4
s
B

T
3
g.t.n1
+

s.1
D
1
(8.125)
where D
1
(m) is the positive distance between the surface and the middle of the rst
soil layer, and T
1.t
(K) is the temperature at the middle of the top soil layer, from
(8.91). Equation (8.125) is iterated four times without updating the saturation
specic humidity in the soil during each iteration.
Finally, the grid-cell averaged uxes of sensible heat (J m
2
s
1
) and moisture
(kg m
2
s
1
) to the boundary layer, accounting for bare and vegetated surfaces,
are
H
f
= H
f
f
v

a
c
p.d
R
a
p
(z
r
)
T
af
P
f
(8.126)
E
f
= E
f
f
v

a
R
a
g
[ q
v
(z
r
) q
af
] (8.127)
respectively (positive is up in both cases), where H
f
and E
f
are the values over bare
soil from (8.105) and (8.106), respectively.
Figure 8.5 shows time series predictions of foliage air temperature (T
af
), foliage
temperature (T
f
), soil-top temperature (

T
g
), and above-canopy air temperature (

T
a
)
over sandy loam in a grid cell of an atmospheric model containing a near-surface
air temperature monitoring site at Lodi, California. The Lodi site is rural and
located in the San Joaquin Valley, California. The Lodi cell contains about 58.4
percent sandy loam and 41.6 percent clay loam. Figure 8.5(a) indicates that during
the day, foliage temperatures exceeded foliage air temperatures, which exceeded
ground temperatures, which exceeded air temperatures. Figure 8.5(b) shows the
modeled differences in temperatures between sandy loam and clay loam in the cell.
Differences in soil-top temperature between the two soil types were up to 10

C
264
10
30
50
70
0 24 48 72 96
T
e
m
p
e
r
a
t
u
r
e

(
o
C
)
Hour after first midnight
20
15
10
5
0
5
10
0 24 48 72 96

T
e
m
p
e
r
a
t
u
r
e

(
o
C
)

(a) (b)
0
10
20
30
40
50
0 24 48 72 96
T
e
m
p
e
r
a
t
u
r
e

(
o
C
)
(c)
Figure 8.5 (a) Time series plot of predicted foliage air temperature (T
af
) , foliage temperature
(T
f
) , soil-top temperature (

T
g
) ---, and air temperature above the canopy (T
a
) over
sandy loam soil in a model grid cell containing the Lodi (LOD) monitoring site. (b) Difference
between temperatures over sandy loam and those over clay loam in the same model grid cell.
(c) Predicted, , versus measured, ----, air temperatures at 5 m at the Lodi monitoring site.
The predicted values were obtained from bilinear interpolation among four grid cells near or
containing the Lodi site. From Jacobson (2001a).
during the day. Figure 8.5(c) compares modeled with measured near-surface air
temperatures at Lodi when temperatures were averaged proportionately over both
soil types in the grid cell.
8.6.4 Road and rooftop surfaces
In this section, temperatures of road and rooftop surfaces are discussed. Construc-
tion materials, such as asphalt, concrete, wood, brick, and composites cover a large
fraction of urban surfaces. Oke et al. (1999) found that 25 percent of surfaces
affecting the energy balance in a densely built-up central Mexico City site con-
sisted of impervious ground material (e.g., roads, sidewalks), 32 percent consisted
of rooftops, 42 percent consisted of walls, and 1 percent consisted of vegetation.
A Vancouver site consisted of 32 percent impervious ground material, 37 percent
rooftops, 27 percent walls, and 4 percent vegetation. Vegetation fractions were
94 percent for a rural site, 70 percent for an urban residential site, and 16 percent
for an urban commercial site.
265
8.6.4.1 Road surfaces
Road surfaces often consist of asphalt or concrete. An average asphalt road is about
6 cm thick and overlies soil. The temperature at the top of an asphalt surface can
be solved iteratively from the equilibrium energy balance equation with
T
g.t.n
=

T
g.t.n1
+

a
c
p.d
R
a
p
(z
r
)
T
g.t.n1
P
g

a
L
e
R
a
d
[ q
v
(z
r
) q
v.s
(

T
g.t.n1
)]
F
s
+

F
i

B
as

T
4
g.t.n1
as
D
1
(T
1.c.t

T
g.t.n1
)

a
c
p.d
R
a
P
g
+4
as
B

T
3
g.t.n1
+

as
D
1
(8.128)
(Jacobson 2001a) where
as
= 0.95 is the thermal-IR emissivity of asphalt (Oke
1978), D
1
(m) is the distance between the ground surface and the middle of the rst
asphalt layer, and
d
is from (8.119), which requires the calculation of W
c
from
(8.124). Here, W
c
accounts for condensation/evaporation and precipitation/runoff
over an asphalt surface. When precipitation or condensation (dew formation)
occurs, all liquid water aboveW
c.max
is treated as runoff, and the liquid water
content below W
c.max
is allowed to evaporate.
Below the asphalt top, temperatures and moisture can be solved with the heat-
conduction and water transport equations, (8.91) and (8.95), respectively. In that
case, model layers, down to 6 cm, may be treated as asphalt, and the remaining
layers as soil containing liquid water. The asphalt is treated as impermeable to
water, thus, the liquid water content of the asphalt is set to zero, and the water
transport equation is solved only under the asphalt base. The heat conduction
equation may be solved assuming a mean thermal conductivity of various asphalts
of
as
= 1.7 W m
1
K
1
(Anandakumar 1999).
8.6.4.2 Rooftops
Many types of roong exist, and rooftop types generally differ between residen-
tial and commercial buildings. In the United States, four out of ve residential
rooftops contain asphalt shingles. Asphalt-shingle roof systems consist of asphalt
shingles overlying saturated felt, overlying a water-impermeable membrane, over-
lying a roof deck, overlying insulation or air. Asphalt shingles are either organic or
berglass based. Organic-based shingles contain a base of cellulose bers, which
is saturated with an asphalt coating and surfaced with weather-resistant mineral
granules. Fiberglass-based shingles contain a base of glass bers surfaced with
an asphalt coating and weather-resistant mineral granules. The saturated felt is
266
10
20
30
40
50
60
70
0 24 48 72 96
T
e
m
p
e
r
a
t
u
r
e

(
o
C
)
0
10
20
30
40
50
0 24 48 72 96
T
e
m
p
e
r
a
t
u
r
e

(
o
C
)
(b) (a)
Figure 8.6 (a) Predicted ground temperatures over four soil types, road surfaces, and rooftops
at Fremont: silt loam, clay, --- loam, road, silt clay loam, roof.
(b) Comparison of modeled, , with measured, ----, near-surface air temperature at Fremont
when all surface types were considered in the model (from Jacobson 2001a).
an asphalt-impregnated, organic-based felt between the roong material and the
waterproof membrane (Asphalt Roong Manufacturers Association 1999). Other
residential rooftops consist of wood shake, concrete tiles, or slate over a membrane
over a deck and insulation over air.
Many commercial rooftops also consist of asphalt shingles, but others consist
of two to four layers of bitumen strengthened with a fabric, such as polyester or
berglass, or with a felt. Gravel or granules are usually embedded on the top layer
of strengthened bitumen. The bitumen layers usually overlie insulation, a vapor
retarder, and a deck. Bitumen is a black or dark-colored cement-like, solid, semi-
solid, or viscous substance composed of high-molecular-weight hydrocarbons and
is found in asphalts, tars, pitches, and asphaltines.
Roof surface temperatures can be solved with (8.128) and subsurface material
temperatures can be solved with the heat conduction equation from(8.91), but with
a different conguration of materials than in the case of road surfaces. One possible
conguration is to assume that the rooftop consists of 1.5 cm of asphalt shingles
over 1 cm of saturated felt (asphalt-felt composite) over 1.25 cm of plywood base
over air. No liquid water is permitted to transmit through or below the roof, and
the air below the roof contains no liquid water, so the water transport equation is
not necessary to solve in this case.
8.6.4.3 Application
Figure 8.6(a) shows modeled soil, road, and roof surface temperatures at Fremont,
California. Roads were the hottest surfaces during the day, since asphalt contains
no water, and energy conduction to soil below asphalt is slow. Peak temperatures
over asphalt varied from 43 to 52

C. Anandakumar (1999) shows a peak August
temperature over an asphalt road at Vienna, Austria, of 44
C. Vienna is 10
latitude
further north than Fremont. Rooftop temperatures in Fig. 8.6 were cooler than
267
road temperatures because air under a rooftop convects energy conducted to it
from the roof base away faster than soil conducts energy conducted to it by the
asphalt base away. Figure 8.6(b) compares modeled with measured near-surface
air temperatures at Fremont when all surface types were considered.
8.6.5 Snow on soil, vegetation, and asphalt
The depth of snowover soil, vegetation, roads, and rooftops is affected by snowfall,
deposition/sublimation, and melting. A prognostic equation for snow depth (m)
accounting for these factors is
D
s.t
= D
s.th
+h

P
s
+h

a
sn
f
s
[ q
v
(z
r
) q
v.s
(min(T
g.t
. T
s.m
))]
R
a
+ f
v
[q
af
q
v.s
(min(T
g.t
. T
s.m
))]
R
f
+h
f
s

a
c
p.d
R
a
p
(z
r
)
T
s.m
P
g
f
v

a
c
p.d
R
f
T
af.t
P
f
T
s.m
P
g
f
s

a
L
s
R
a
[ q
v
(z
r
) q
v.s
(T
s.m
)] f
v

a
L
s
R
f
[q
af
q
v.s
(T
s.m
)]
f
s

F
s

F
i
+
B
sn
T
4
s.m

sn
D
1
(T
1.t
T
s.m
)
sn
L
m
(8.129)
where

P
s
is the snowfall rate (ms
1
), L
m
is the latent heat of melting (J kg
1
),
L
s
is the latent heat of sublimation (J kg
1
), q
v.s
(T
s.m
) is the saturation specic
humidity (kg kg
1
) over ice at the melting point of snow, T
s.m
= 273.15 K,
sn
is
the mass density of snow (kg m
3
),
sn
= 0.99 is the emissivity of snow, and
sn
=
0.08 W m
1
K
1
(Oke 1978) is the thermal conductivity of snow. The rst term in
the equation accounts for snowfall, the second accounts for deposition/sublimation,
and the third accounts for melting. The melting term is applied only if the term is
less than zero (energy is added to the system) and the snow temperature exceeds
the melting point of snow.
The temperature at the top of a snow surface is calculated with (8.125) or
(8.128), except (a)
g
= 1 and
d
= 1, (b) q
v.s
is dened as the saturation specic
humidity over ice, not liquid water, (c) L
s
replaces L
e
, (d)
sn
is used instead of
s
,
v
, or
as
, (e)
sn
is used instead of
s
or
as
, and (f) transpiration through leaf
surfaces is set to zero (E
t
= 0).
8.6.6 Water, sea ice, and snow over sea ice
Over oceans, inland seas, and lakes, water-top temperature is affected by sensible,
latent, and radiative heat uxes as well as energy advection (due to water circula-
tion). In the mixed layer of the ocean, which is its top layer of depth D
l
(typically
50100 m), in which energy and salinity are well mixed, the water temperature
268
can be calculated noniteratively in time with
T
g.t
=

T
g.th
+h

a
c
p.d
R
a
p
(z
r
)
T
g.th
P
g

a
L
e
R
a
[ q
v
(z
r
) q
v.s
(

T
g.th
)]
F
s
+

F
i

B
w

T
4
g.th
sw
c
p.sw
D
l
(8.130)
where q
v.s
(T
g.th
) (kg kg
1
) is the saturation specic humidity over liquid water at
the temperature of the water,
w
=0.97 is the thermal-IR emissivity of liquid water
(Oke 1978),
sw
is the density of seawater (1028 kg m
3
at 0

C in the presence of
35 parts per 1000 of salinity; Lide 2003), and c
p.sw
is the temperature-dependent
specic heat of liquid water at constant pressure (3986.5 J kg
1
K
1
at 0

C in the
presence of 35 parts per 1000 of salinity).
Equation (8.130) accounts for all energy uxes to and from the ocean mixed
layer except for horizontal advective uxes and diffusive uxes to the deep ocean.
Such uxes need to be calculated with an ocean circulation model, which solves
equations similar to the equations of atmospheric dynamics, but for ocean water.
Circulation models can be two- or three-dimensional. The two-dimensional models
resolve the mixed layer only. The three-dimensional models account for transfer
between the mixed layer and deep ocean as well.
When the water temperature falls below the freezing point of seawater in the
presence of 35 parts per 1000 of salinity, T
i.f
=271.23 K(Lide 2003), sea ice begins
to form on top of ocean water. The sea ice surface temperature can be determined
with the four-iteration surface-energy balance calculation,
T
g.t.n
= T
g.t.n1
+

a
c
p.d
R
a
p
(z
r
)
T
g.t.n1
P
g

a
L
s
R
a
[ q
v
(z
r
) q
v.s
(

T
g.t.n1
)]
F
s
+

F
i

B
i

T
4
g.t.n1
i
D
i.th
(T
i.f

T
g.t.n1
)

a
c
p.d
R
a
P
g
+4
i

T
3
g.t.n1
+

i
D
i.th
(8.131)
where q
v.s
(T
g.t.n1
) is the saturation specic humidity over ice,
i
= 0.97 is the
thermal-IR emissivity of ice, and
i
= 2.20 W m
1
K
1
is the thermal conductivity
of ice (Lide 2003). The change in sea ice thickness (m) at the sea ice/air interface
269
due to sublimation/deposition and melting is
D
i.t
= D
i.th
+h

a
i
R
a
[ q
v
(z
r
) q
v.s
(min(

T
g.t
. T
i.m
))]
+h

a
c
p.d
R
a
p
(z
r
)
T
i.m
P
g

a
L
s
R
a
[ q
v
(z
r
) q
v.s
(T
i.m
)]

F
s

F
i
+
B
i
T
4
i.m

i
D
i.th
(T
i.f
T
i.m
)
i
L
m
(8.132)
where
i
(kg m
3
) is the density of ice (916.7 kg m
3
) at its melting point,
T
i.m
= 273.05 K. The melting (second) term applies only if the term is less than
zero (energy is added to the system) and if the temperature from (8.131) exceeds
the melting point.
The change in sea ice thickness due to freezing/melting at the water/sea ice
interface is
D
i.t
= D
i.th
+h
F
b

i
D
i.th
(T
i.f

T
g.t
)
i
L
m
(8.133)
where

F
b
(W m
2
) is the net ux of energy from below the interface, estimated
as 2 W m
2
for Arctic waters and 25 W m
2
for Antarctic waters (Parkinson
and Washington 1979), and the second term accounts for conduction between the
interface and the top of the ice. Maximum ice thicknesses are typically 4 m in the
Arctic and 1.5 m in the Antarctic.
When sea ice exists and snowfall occurs, the snowcovers the ice. The temperature
at the top of the snow overlying the sea ice layer can be modeled with
T
g.t.n
=

T
g.t.n1
+

a
c
p.d
R
a
p
(z
r
)
T
g.t.n1
P
g

a
L
s
R
a
[ q
v
(z
r
) q
v.s
(

T
g.t.n1
)]
F
s
+

F
i

B
sn

T
4
g.t.n1
sn
sn
D
i.th
+
i
D
s.th
(T
i.f

T
g.t.n1
)

a
c
p.d
R
a
P
g
+4
sn

T
3
g.t.n1
+

sn
sn
D
i.th
+
i
D
s.th
(8.134)
where the thermal conductivity termaccounts for conduction through ice and snow
layers. An estimate of the temperature at the snowice interface is
T
I
=

sn
D
i.th

T
g.t
+
i
D
s.th
T
i.f
sn
D
i.th
+
i
D
s.th
(8.135)
(Parkinson and Washington 1979).
270
8.8 Problems
8.7 SUMMARY
In this chapter, boundary-layer and surface processes were discussed. Expressions
for kinematic turbulent uxes of momentum, energy, and moisture were derived.
For the surface layer, such uxes were described in terms of bulk aerodynamic for-
mulae and MoninObukhov similarity theory. Some similarity-theory parameters
are friction wind speed, the potential temperature scale, and the MoninObukhov
length. Fluxes of momentum, energy, and moisture above the surface layer were
parameterized with a hybrid model and with a high-order turbulence closure tech-
nique. Equations describing soil temperature and moisture content were then dis-
cussed. Ground temperatures depend on the specic heat, thermal conductivity,
and moisture content of soil and on uxes of radiative, sensible, and latent heat
between the soil and air. Prognostic equations for temperature and moisture over
soil, vegetation-covered soil, rooftops, road surfaces, water, sea ice, and snow over
all surfaces were described.
8.8 PROBLEMS
8.1 Calculate the surface roughness length for momentum over a rough ocean
assuming u
= 0.3 m s
1
and over vegetated land assuming a forest canopy
with top h
c
= 20 m and L
T
= 7 m
2
m
2
. Calculate the wind speed 100 m
above the ocean surface and above h
c
assuming a displacement height of
d
c
= 15 m for the canopy. Assume a neutral boundary layer and logarithmic
wind prole in both cases. In which case is the wind speed higher. Why?
8.2 If conditions at the surface are z
r
= 10 m, z
0.m
= 0.05 m, z
0.h
= 0.0005 m,
u(z
r
) = 8 ms
1
, : (z
r
) = 2 ms
1
,

T (z
r
) = 285 K,

T(z
0.h
) = 286 K, p
a
(z
r
) =
1004 hPa, and p
a
(z
0.h
) = 1005 hPa, calculate u
, L, Ri
b
, and K
m.zx
at the
reference height. Assume dry air.
8.3 Assume z
0.m
= 0.01 m, u
= 1 m s
1
, u(z
r
) = 10 m s
1
, and : (z
r
) = 5 m s
1
at z
r
= 10 m. Calculate the eddy diffusion coefcient for momentum,
K
m.zx
.
8.4 Assuming u
= 0.1 m s
1
, p
a
= 998 hPa, and the air is dry, calculate the
surface roughness lengths for energy and water vapor when

T = 288 K.
8.5 Compare vertical turbulent sensible-heat uxes over the ocean from bulk
aerodynamic formulae and MoninObukhov similarity theory. Assume the
same conditions as in Problem8.2, except assume z
0.m
=0.00001 m. Assume
v
at the roughness length for energy equals that at the roughness length for
momentum. Discuss differences in results.
8.6 Using conditions from Problems 8.2 and 8.5, calculate the eddy diffusion
coefcient for energy with bulk aerodynamic formulae and similarity theory.
Discuss differences in results.
8.7 Assume all conditions are the same as in Problem8.2, except that the relative
humidity at the reference height is now85 percent, and the specic humidity
at the roughness length for moisture equals the saturation specic humidity
at the roughness length for energy. Estimate the vertical turbulent ux of
moisture from similarity theory.
271
8.8 Calculate the thermal conductivity of a soilwaterair mixture for a sandy-
loam soil when the volumetric water content is 0.2 and 0.05 m
3
m
3
. In
which case is the mixture more conductive? Why?
8.9 Simplify (8.130) to calculate the temperature of the mixed layer of the ocean
inthe absence of sensible andlatent heat uxes andassuming the only source
or sink of thermal-infrared radiation is the sink due to upward emission
(
B
w
T
4
g.th
). Calculate the temperature after two hours if the initial temper-
ature is 290 K, the downward solar irradiance is 800 W m
2
, and the thick-
ness of the mixed layer is 70 m. Should the temperature change increase or
decrease if the mixing depth increases?
8.10 If z
r
=10 m, u(z
r
) =3 ms
1
, :(z
r
) =16 ms
1
,

T(z
r
) =293 K,

T(z
0.h
) =292.5 K,
p
a
(z
r
) = 1002 hPa, and p
a
(z
0.h
) = 1003 hPa, write a script to calculate u
,
L, Ri
b
, and K
m.zz
as a function of z
0.m
when the air is dry. Calculate values
for 0.000 01 m - z
0.m
- 5m. Assume z
0.h
z
0.m
100. Plot the results for each
variable.
8.11 Write a computer script to calculate the change of wind speed and potential
virtual temperature with height in the surface layer from similarity theory.
Assume the same conditions as in Problem 8.2. Plot the results from the
ground surface up to 10 m altitude. Change

T(z
r
) to 286 K, and replot the
results. Discuss the differences between the two cases.
8.12 Write a computer program to calculate the temperature over a single asphalt
layer 6 cmthick from(8.128) assuming air temperature at the reference height
(10 m) is a constant 295 K, the initial ground temperature is 285 K, the con-
stant soil temperature below the ground is 290 K, the surface air pressure
is 1013 hPa, the pressure at the reference height is 1012 hPa, the incident
solar radiation varies with a sine function between 06:00 and 18:00, with a
peak value of 1000 W m
2
at 12:00 and zero values before 06:00 and after
18:00, and the downward thermal-IR irradiance is zero. In addition, assume
R
a
= 0.74 ln(z
r
z
0.h
)ku
, where z
0.h
= 0.000 135 m and u
= 0.4 m s
1
.
Finally, assume q
v
(z
r
) = 0. Use the program to calculate the ground tem-
perature from 03:00 to 21:00 on a single day. Plot and discuss the result.
(Hint: take a time step of 1 s. Iterate the equation four times for each time
step. At the end of each time step, update temperatures and solar radiation
on the right side of the equation).
272
9
Radiative energy transfer
R
adi ati on through the atmosphere affects temperature, pollutant concen-
tration, visibility, and color. Temperature is affected by heating and cooling
fromultraviolet, visible, and infrared radiation interactions with the ground, gases,
aerosol particles, and hydrometeor particles. Pollution is affected by interactions of
ultraviolet and some visible radiation with gases. Visibility and colors are affected
by interactions of visible radiation with gases, aerosol particles, and hydrometeor
particles. In this chapter, radiation laws, optical properties of gases and particles,
light processes, and a solution to the radiative transfer equation are discussed.
The radiative transfer equation is used to determine rates of heating, cooling, and
molecular photolysis. Important radiative laws discussed here include Plancks law,
Wiens law, and the StefanBoltzmann law. Equations quantifying the level of atten-
uation and redirection of radiation by gases, particles, and cloud drops are also
described.
9.1 ENERGY TRANSFER PROCESSES
Radiation is the emission or propagation of energy in the form of a photon or elec-
tromagnetic wave. A photon is a particle or quantum of electromagnetic energy
that has no mass, no electric charge, and an indenite lifetime. An electromag-
netic wave is a disturbance traveling through a medium, such as air or space, that
transfers energy from one object to another without permanently displacing the
medium itself. Electromagnetic waves may be considered as dual transverse waves
in that they consist of an electric wave and a magnetic wave in phase with and at
right angles to each other and to the direction of propagation. Radiation is emitted
by all bodies in the Universe that have a temperature above absolute zero (0 K).
Once emitted, radiation passes through space or air to another body. Upon reach-
ing the second body, the radiation can be reected, scattered, absorbed, refracted,
dispersed, or transmitted. Each of these processes is discussed in this chapter.
When a body, such as air, emits more radiation than it absorbs, its tempera-
ture decreases. When a body absorbs more radiation than it emits, its temperature
increases. Other processes that affect air temperatures include advection, forced
convection, turbulence, and latent-heat exchange. The thermodynamic energy
equation in (3.76) takes these processes into account.
Figure 9.1 shows the relative importance of several processes on the atmospheric
energy budget. Solar radiation provides energy for atmospheric, cloud, and surface
273
Sun
Water
Evaporation
(latent-heat transfer)
4 6
100
16
4
7
50
20
34
6
30
50
30
66
60
23
Atmosphere
Thermal / mechanical
turbulence (sensible-heat transfer)
I
n
f
r
a
r
e
d

r
a
d
i
a
t
i
o
n
S
o
l
a
r

r
a
d
i
a
t
i
o
n
Figure 9.1 Energy balance for Earthatmosphere system.
Values are dimensionless relative quantities of energy. The
sum of sources minus sinks for clouds, the atmosphere, or
the Earth equals zero. For example, clouds are in radiative
balance, since they absorb and emit 64 units of radiation.
90 60 30 0 30 60 90
R
a
d
i
a
n
t

e
n
e
r
g
y

p
e
r

y
e
a
r
Latitude
Surplus
Deficit
Deficit
Absorbed incoming solar
Emitted
outgoing infrared
Transfer
Figure 9.2 Schematic showing that a glob-
ally and yearly averaged energy surplus at
the Equator and decit at the poles is com-
pensated for by energy transfer from the
Equator toward the poles.
heating. The Earths surface emits infrared radiation, much of which is absorbed by
greenhouse gases and clouds and some of which escapes to space. The atmosphere
and clouds emit infrared radiation in all directions. Turbulence transfers energy
from the surface to the troposphere. Evaporation from the surface releases water
vapor, which stores latent heat. The water vapor travels to the free troposphere,
where it may condense to form clouds, releasing the stored latent heat.
The transport of latent heat via water vapor is an important process by which
solar energy absorbed near the Equator is transferred poleward. Figure 9.2 shows
that, at the Equator, the surface absorbs more solar radiation than it emits infrared
274
9.2 Electromagnetic spectrum
radiation, causing an energy surplus. At the poles, the reverse is true, and an energy
decit occurs. In the absence of energy transfer between the Equator and poles, tem-
peratures near the poles would continuously decrease, and those near the Equator
would continuously increase. On Earth, energy is continuously supplied from the
Equator to the poles by three processes. One is poleward transport of energy by
winds, the second is poleward transport of energy by ocean currents, and the third
is the poleward transport of water vapor and its stored latent heat by winds. When
the water vapor condenses, mostly at midlatitudes, it releases its latent heat to the
air around it.
9.2 ELECTROMAGNETIC SPECTRUM
Whether radiation is considered an electromagnetic wave or photon, it travels at
the speed of light. If radiation propagates as a wave, its wavelength is
=
c
=
1

(9.1)
where is the waves frequency of oscillation, or number of wavelengths that pass
through a point per unit time (s
1
), c is the speed of light (e.g., m s
1
) and = 1
is the wavenumber (e.g., m
1
), dened here as the number of wavelengths per unit
distance. This denition of wavenumber, applied here to electromagnetic waves,
differs from that given in (4.89), which was applied to atmospheric waves. Units
for wavelength are micrometers (1 m = 10
6
m), nanometers (1 nm = 10
9
m),
centimeters (cm), or meters (m). In a vacuum, the speed of light is c = 2.9979
10
8
m s
1
.
In 1900, Max Planck theorized that the total radiative energy (J) emitted or
absorbed by a substance was
E
t
= nh = n
hc
(9.2)
where n is an integer, called a quantum number, and h = 6.6256 10
34
J s is
Plancks constant. Equation (9.2) states that substances do not emit or absorb
radiative energy continuously, but in packets or quanta. Although Planck believed
that light was emitted or absorbed discontinuously, he held that it traveled through
space as an electromagnetic wave. In 1905, Albert Einstein postulated that light
energy traveled through space in concentrated bundles, and the energy of each
bundle was
E
p
= h =
hc
(9.3)
In 1926, Gilbert Lewis termed these bundles photons. Equation (9.3) represents the
energy emitted or absorbed per photon (J photon
1
) and depends on the wavelength
(or frequency) of radiation. The greater the frequency, the shorter the wavelength,
275
and the greater the energy of a quantum. The electromagnetic and photon theories
of light are interrelated by (9.2).
Example 9.1
Find the energy emitted per photon, the frequency, and the wavenumber of a
= 0.5-m and = 10-m wavelength of energy.
SOLUTION
When = 0.5 m, E = 3.97 10
19
J photon
1
, = 5.996 10
14
s
1
, and =
2 m
1
. When = 10 m, E = 1.98 10
20
J photon
1
, = 2.998 10
13
s
1
,
and = 0.1 m
1
. Shorter wavelengths generate more energy per photon than
do longer wavelengths.
From(9.2), Planck derived an equation relating the intensity of radiant emission
from a perfectly emitting substance to the absolute temperature of the substance
and the wavelength of emission. A perfectly emitting substance is one that, in
thermodynamic equilibrium, emits all radiation that it absorbs. Absorption occurs
when electromagnetic energy enters a substance and is converted to internal energy.
If a bodys emission is less than its absorption, the body is out of equilibrium, and
its temperature rises.
A blackbody is a substance that absorbs all radiation that is incident upon it.
No incident radiation is reected by a blackbody. No bodies are true blackbodies,
although the Earth and the Sun are close, as are black carbon, platinum black,
and black gold (e.g., Siegel and Howell 1992). The term blackbody was coined
because good absorbers of visible radiation generally appear black. However, good
absorbers of infrared radiation are not necessarily black. For example, one such
absorber is white oil-based paint.
Blackbodies not only absorb, but also emit the maximum possible intensity of
radiant energy at a given wavelength and temperature. Planck determined this
intensity as
B
.T
=
2hc
2
exp
hc
k
B
T
1
(9.4)
now called Plancks law, where B
.T
is radiant intensity or radiance (Wm
2
m
1
sr
1
), and k
B
23
J K
1
=W s K
1
). Radiance is
the energy emitted per unit area per unit time per unit wavelength per incremental
solid angle (units of steradians). Since the radiance in (9.4) is dened for individual
wavelengths, it is called a spectral radiance. Incremental solid angle is dened
shortly.
276
Table 9.1 Emissivities of different surface types for a typical infrared wavelength
Surface type Emissivity (fraction) Surface type Emissivity (fraction)
Liquid water 1.0
a
Soil 0.90.98
b
Fresh snow 0.99
b
Grass 0.90.95
b
Old snow 0.82
b
Desert 0.840.91
b
Liquid water clouds 0.251.0
c
Forest 0.950.97
b
Cirrus clouds 0.10.9
c
Concrete 0.710.9
b
Ice 0.96
d
Urban 0.850.87
a
a
Seaman et al. (1989),
b
Oke (1978),
c
Liou (2002),
d
Sellers (1965).
Equation (9.4) applies to any blackbody with a temperature above absolute zero
(0 K). Real substances are generally not perfect emitters. Instead, they usually emit
a fraction of the radiance that a blackbody emits. The radiance (Wm
2
m
1
sr
1
)
actually emitted by any substance is approximately
e
B
.T
(9.5)
where
(dimensionless) is the emissivity of the substance. Emissivity is the frac-

tion (1) of B
.T
actually emitted. Emissivities depend on wavelength. Table 9.1
gives emissivities for some surface types in the infrared part of the electromagnetic
spectrum.
Absorptivity (a
) is the fraction (1) of incident radiation that a substance actu-

ally absorbs. Kirchoffs law (named for Gustav Kirchoff, 182487) states that, in
thermodynamic equilibrium, the emissivity and absorptivity of a substance equal
each other (a
). Thus, the efciency at which a substance absorbs radia-

tion equals that at which it emits radiation, and a perfect emitter of radiation
(
= 1) is also a perfect absorber of radiation (a
= 1) and a blackbody. Suppose

a blackbody object is placed in a vacuum enclosed by blackbody walls. Over time,
the temperatures of the object and walls equalize. If Kirchoffs law did not hold
at this point, a net heat transfer would occur between the object and the walls
even though the object and wall temperatures are the same. Such a heat transfer
violates the second law of thermodynamics. Thus, the absorptivity and emissivity
of a substance must equal each other in equilibrium.
An incremental solid angle, dO
a
, used in the denition of radiance, is an incre-
mental surface area on a unit sphere, which is a sphere with radius normalized to
unity. The equation for incremental solid angle is
dO
a
=
dA
s
r
2
s
(9.6)
where dA
s
is an incremental surface area, and r
s
is the radius of a true sphere.
Incremental solid angle has units of steradians (sr), which is analogous to units of
radians for a circle.
The incremental surface area in (9.6) can be found from Fig. 9.3. In the gure,
a line is drawn from the center of the sphere to the center of an incremental area
277
dA
s
r
s
sin df
r
s
r
s
d
z
x
y
O
Figure 9.3 Radiance, emitted from
point (O) on a horizontal plane,
passes through an incremental area
dA
s
at a distance r
s
from the point
of emission. The angle between the
z-axis and the angle of emission is
the zenith angle (), and the hor-
izontal angle between a reference
axis (x-axis) and the line of emis-
sion is the azimuth angle (). The
size of the incremental surface area
is exaggerated.
dA
s
, which is a distance r
s
from the spheres center. The line is directed at a zenith
angle from the surface normal (where the surface is on the xy plane). The line
is also located at an azimuth angle , directed counterclockwise from the positive
x-axis to a horizontal line dropped from the line. From the geometry shown, the
incremental surface area is
dA
s
= (r
s
d)(r
s
sin d) = r
2
s
sin d d (9.7)
where d and d are incremental zenith and azimuth angles, respectively. Substi-
tuting (9.7) into (9.6) gives the incremental solid angle as
dO
a
= sin d d (9.8)
Integrating dO
a
over all possible solid angles around the center of a sphere gives
the solid angle around a sphere as
O
a
=
O
a
dO
a
=
2
0

0
sin d d = 4 steradians (9.9)
The solid angle around the center of the base of a hemisphere is 2 steradians.
Plancks law gives the spectral radiance, B
.T
(W m
2
m
1
sr
1
) emitted by
a blackbody. In terms of Fig. 9.3, imagine that the xy plane is the surface of an
object. The radiance emitted from point O can travel in any direction above the
xy plane. If it travels through area dA
s
, it passes through an incremental solid
278
angle, dO
a
, which is through a cone originating at point O. B
.T
is the spectral
radiance emitted by a blackbody at the bodys surface. Radiance changes with
distance through space. I
is dened here as the spectral radiance at a given point

in space, regardless of the source of the radiance. At the surface of a blackbody,
I
= B
.T
.
In atmospheric models, two quantities related to spectral radiance are commonly
calculated. These are spectral actinic ux and spectral irradiance (e.g., Madronich
1987). Spectral actinic ux (also called spherical intensity or actinic irradiance)
is the integral of spectral radiance over all solid angles of a sphere. It is used for
calculating photolysis coefcients of gases in the atmosphere. Since gas molecules
can absorb radiation, regardless of the direction the radiation originates from,
radiance is integrated over a sphere to obtain actinic ux.
Spectral actinic ux is determined by integrating incremental actinic ux over
all solid angles around the center of a sphere. The incremental actinic ux is
dE
= I
dO
a
(9.10)
where E
is in units of W m
2
m
1
but is typically converted to units of pho-
tons cm
2
s
1
m
1
for photolysis calculations (Section 9.8.4). The integral of
incremental actinic ux over a sphere is
E
O
a
dE
O
a
I
dO
a
=
2
0

0
I
sin d d (9.11)
When radiation propagates with equal intensity in all directions, I
is independent
of direction. In such cases, the radiance is called isotropic, and (9.11) simplies to
E
= I
2
0

0
sin d d = 4I
(9.12)
Equation (9.12) states that the isotropic spectral actinic ux equals 4 multiplied
by the spectral radiance, where the units of 4 are steradians.
Spectral irradiance (also called the net ux or energy ux) is the vertical com-
ponent of spectral radiance, originating from all directions above a at plane, that
passes across the plane, per unit surface area, time, and wavelength. In terms of
Fig. 9.3, it is the vertical component of radiance, integrated over the hemisphere
above the xy plane. For a zenith angle of = 0
, the irradiance impinging on the

xy plane is maximum. For a zenith angle of = 90
, the irradiance impinging on

the xy plane is zero.
Spectral irradiance is calculated by integrating the component of radiance nor-
mal to the xy plane over all solid angles of the hemisphere above the xy plane. If
I
is the spectral radiance passing through point Oin Fig. 9.3, I
cos is the compo-

nent of radiance normal to the xy plane. Multiplying this quantity by incremental
solid angle gives the incremental spectral irradiance normal to the xy plane as
dF
= I
cos dO
a
(9.13)
279
10
24
10
16
10
8
10
0
10
8
10
16
10
24
10
32
10
40
10
6
10
4
10
2
10
0
10
2
10
4
10
6
10
8
10
10
10
12
10
6
10
4
10
2
10
0
10
2
10
4
10
6
10
8
10
10
10
12
R
a
d
i
a
t
i
o
n

i
n
t
e
n
s
i
t
y

(
W

m
1
)
Wavelength (m)
6
0
0
0

K
1
5

m
i
l
l
i
o
n

K
G
a
m
m
a
X
U
V
V
i
s
i
b
l
e
I
n
f
r
a
r
e
d
S
h
o
r
t

r
a
d
i
o
A
M

r
a
d
i
o
L
o
n
g

r
a
d
i
o
T
e
l
e
v
i
s
i
o
n

&

F
M

r
a
d
i
o
3
0
0

K
1

K
Figure 9.4 Spectral irradiance as a function of wave-
length and temperature resulting from emission by a
blackbody. Data from (9.16) and (9.4).
(W m
2
m
1
). Integrating (9.13) over the hemisphere above the xy plane in
Fig. 9.3 gives
F
O
a
dF
O
a
I
cos dO
a
=
2
0
2
0
I
cos sin d d (9.14)

For isotropic emission, (9.14) simplies to
F
= I
2
0
2
0
cos sin d d = I
(9.15)
Thus, the isotropic spectral irradiance is multiplied by the isotropic radiance,
where units of are steradians. At the surface of a blackbody, the spectral irradiant
emission is
F
= I
= B
.T
(9.16)
Figure 9.4 shows irradiant emission from a blackbody versus wavelength for
different temperatures, obtained from(9.16) and (9.4). The gure shows that hotter
bodies emit much more energy than do cooler bodies.
The sources of irradiant energy in the Earths atmosphere are the Sun, the surface
of the Earth, and the atmosphere itself. The temperature at the center of the Sun
is about 15 million K, and the Sun emits primarily gamma rays and X-rays, as
shown in Fig. 9.4. Most of these waves are redirected or absorbed during their
random walk toward the outside of the Sun. Solar radiation that reaches the Earth
originates mainly from the visible surface of the Sun, the photosphere, which has
an effective temperature near 6000 K (closer to 5800 K). This temperature results
in the solar spectrum of radiation shown in Fig. 9.5. The solar spectrum includes
280
10
0
10
2
10
4
10
6
10
8
0.01 0.1 1 10 100
R
a
d
i
a
t
i
o
n

i
n
t
e
n
s
i
t
y

(
W

m
1
)
Wavelength (m)
Sun
Earth
Ultraviolet
Visible
Infrared
Figure 9.5 Irradiant emission versus wave-
length for the Sun and the Earth when both are
considered perfect emitters. Data from (9.16)
and (9.4).
0 10
0
2 10
7
4 10
7
6 10
7
8 10
7
1 10
8
1.2 10
8
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8
R
a
d
i
a
t
i
o
n

i
n
t
e
n
s
i
t
y

(
W

m
1
)
Wavelength (m)
U
V
-
C
U
V
-
B
U
V
-
A
Visible
Far
UV
Near
UV
R
e
d
G
r
e
e
n
B
l
u
e
Figure 9.6 Ultraviolet and visible portions of the
solar spectrum. Data from (9.16) and (9.4).
ultraviolet, visible, and infrared wavelength regions. The ultraviolet region consists
of wavelengths -0.38 m, the infrared region consists of wavelengths >0.75 m,
and the visible region consists of wavelengths 0.380.75 m. The infrared spectrum
is divided into the near- or solar-infrared (0.754 m) and the far- or thermal-
infrared (>4 m) spectra. Most infrared radiation from the Sun that reaches the
Earth is solar-infrared radiation.
Figure 9.5 shows the irradiant emission spectrumof the Earth. The effective tem-
perature at the surface of Earth is approximately 288 K; thus, almost all radiation
emission from the Earth is thermal-infrared radiation. Even in the stratosphere,
where temperatures can drop to below 200 K, and in hot deserts, where tem-
peratures can exceed 315 K, emission is thermal-infrared. Thus, with respect to
studies of the Earths atmosphere, incoming wavelengths of interest are ultravio-
let, visible and solar-infrared, and outgoing wavelengths of interest are thermal-
infrared.
Figure 9.6 shows the ultraviolet (UV) and visible portions of the solar spec-
trum. The UV spectrum is divided into the far-UV (0.01 to 0.25 m) and near-UV
(0.25 to 0.38 m) spectra, as shown in Fig. 9.6. The near-UV spectrum is further
281
10
0
10
2
10
4
10
6
10
8
0.01 0.1 1 10 100
S
p
e
c
t
r
a
l

i
r
r
a
d
i
a
n
c
e

(
W

m
1
)
Wavelength (m)
6000 K
4000 K
2000 K
1000 K
300 K
0.5 m
10 m
Figure 9.7 Radiation spectrum as a function of
temperature and wavelength. Data from (9.16)
and (9.4). The line through the peaks was
obtained from Wiens law.
divided into UV-A (0.32 to 0.38 m), UV-B (0.29 to 0.32 m), and UV-C (0.25
to 0.29 m) wavelengths. All solar radiation below 0.28 m is absorbed by the
Earths atmosphere above the troposphere. Nitrogen gas (N
2
) absorbs wavelengths
less than 0.1 m in the thermosphere, and oxygen absorbs wavelengths less than
0.245 m in the thermosphere, mesosphere, and stratosphere. Ozone in the strato-
sphere and troposphere absorbs much radiation from 0.17 to 0.35 m. A fraction
of UV-A, UV-B, and UV-C radiation longer than 0.28 m passes through the
stratosphere and troposphere to the ground.
UV-A radiation (tanning radiation) is relatively harmless to most humans.
UV-B radiation can cause sunburn. UV-C radiation can cause severe damage to
many forms of life when received in high dosages. Since ozone in the stratosphere
absorbs most UV-B and UV-Cradiation, ozone reduction in the stratosphere would
allow more UV-B and UV-C radiation to reach the surface, damaging exposed life
forms.
The visible spectrum consists of wavelengths corresponding to the colors of the
rainbow. For convenience, visible light is divided into blue (0.380.5 m), green
(0.50.6 m), and red (0.60.75 m) light.
The peak radiation wavelength emitted at a given temperature can be found from
Wiens displacement law (Wiens law). This law is derived by differentiating (9.4)
with respect to wavelength at a constant temperature and setting the derivative
equal to zero. The result is
p
(m)
2897
T(K)
(9.17)
where
p
is the peak wavelength of emission from a blackbody. The law states that
the hotter a body, the shorter the peak wavelength of radiation emitted. Figure
9.7 shows how peak wavelengths of radiation at different temperatures can be
connected by a line on a loglog scale, where the line is described by (9.17).
282
9.3 Light processes
Example 9.2
In the Suns photosphere, the peak wavelength of emission is near
p
=
28975800 = 0.5m, which is in the visible part of the solar spectrum. At the
Earths surface, the peak wavelength of emission is
p
= 2897288 = 10.1m,
whichis inthe infraredpart of the spectrum. Bothpeak wavelengths are shown
in Fig. 9.7.
Integrating the spectral irradiance in (9.16) over all wavelengths gives the Stefan
Boltzmann law,
F
b
=

0
B
.T
d =
B
T
4
(9.18)
where F
b
is the total irradiance emitted by a blackbody at a given temperature and
B
= 2k
4
B
4
15h
3
c
2
= 5.67 10
8
W m
2
K
4
is the StefanBoltzmann constant.
F
b
is the area under each curve shown in Figs. 9.49.7.
Example 9.3
For an effective temperature of T = 5800 K, the irradiance emitted
by the photosphere is F
b
= 64 10
6
W m
2
. For an effective temperature of
T = 288 K, the irradiance emitted by the Earths surface is F
b
= 390 W m
2
.
9.3 LIGHT PROCESSES
Absorption is one process that affects electromagnetic radiation in the Earths atmo-
sphere. Other processes include reection, refraction, dispersion, diffraction, par-
ticle scattering, and gas scattering. These processes are discussed here.
9.3.1 Reection
Reection occurs when a wave or photon of radiation is absorbed by an object and
reemitted at an angle (called angle of reection) equal to the angle of incidence.
No energy is lost during absorption and reemission. Figure 9.8 shows an example
of reection.
The reectivity of a surface is called its albedo, which is the fraction of sun-
light incident on a surface that is reected. Albedos are wavelength-dependent.
Table 9.2 gives albedos, averaged over the visible spectrum, for several surface
types. The table shows that the albedo of the Earth and atmosphere together (plan-
etary albedo) is about 30 percent. Two-thirds of Earths surface is covered with
water, which has an albedo of 520 percent (typical value of 8 percent), depending
283
Table 9.2 Albedos averaged over the visible spectrum for several
surface types
Surface type Albedo (fraction) Surface type Albedo (fraction)
Earth and atmosphere 0.3
a
Soil 0.050.2
b
Liquid water 0.050.2
b
Grass 0.160.26
c
Fresh snow 0.750.95
d
Desert 0.200.40
b
Old snow 0.40.7
d
Forest 0.100.25
b
Thick clouds 0.30.9
b
Asphalt 0.050.2
c
Thin clouds 0.20.7
b
Concrete 0.10.35
c
Sea ice 0.250.4
b
Urban 0.10.27
c
a
Liou (2002),
b
Hartmann (1994),
c
Oke (1978),
d
Sellers (1965).
Water
q
1
q
3
q
2
Air
Incident Reflected
Refracted
Figure 9.8 Examples of reection and
refraction. During reection,
1
=
3
.
During refraction, the angles of incidence
and refraction are related by Snells law.
largely on the angle of the Sun. Soils and forests also have lowalbedos. Much of the
Earthatmosphere reectivity is due to clouds and ice, which have high albedos.
UV-B albedos are much smaller than are visible albedos over grassland and other
nonsnowsurfaces, but UV-B albedos are slightly larger than are visible albedos over
snow (Blumthaler and Ambach 1988). Harvey et al. (1977) found the UV-AUV-B
albedo to be 3 percent for water and 0.8 percent for grassland.
9.3.2 Refraction
Refraction occurs when a wave or photon travels through a medium of one density
then bends as it enters a mediumof another density. During the transition, the speed
of the wave changes, changing the angle of the incident wave relative to a surface
normal, as shown in Fig. 9.8. If a wave travels from a medium of one density to
a medium of a higher density, it bends (refracts) toward the surface normal. The
angle of refraction is related to the angle of incidence by Snells law,
n
2
n
1
=
sin
1
sin
2
(9.19)
284
9.3 Light processes
Table 9.3 Real indices of refraction of air and liquid water versus wavelength
Wavelength (m) n
air
n
water
Wavelength (m) n
air
n
water
0.2 1.000324 1.396 1.0 1.000274 1.327
0.3 1.000292 1.349 4.0 1.000273 1.351
0.4 1.000283 1.339 7.0 1.000273 1.317
0.5 1.000279 1.335 10.0 1.000273 1.218
0.6 1.000277 1.332 20.0 1.000273 1.480
0.7 1.000276 1.331
Data for air were obtained from (9.21), and data for liquid water were obtained from Hale and
Querry (1973).
In this equation, n is the real index of refraction (dimensionless), is the angle
of incidence or refraction, and subscripts 1 and 2 refer to incident and refracted
light, respectively. The real index of refraction is the ratio of the speed of light in a
vacuum (c) to that in a different medium (c
1
). Thus,
n
1
= cc
1
(9.20)
Since light cannot travel faster than its speed in a vacuum, the real index of refrac-
tion of a medium other than a vacuum must exceed unity. The index of refraction
is wavelength dependent. The real index of refraction of air as a function of wave-
length is approximately
n
a.
1 = 10
8
8342.13 +
2 406 030
130
2
+
15 997
38.9
2
(9.21)
(Edlen 1966) where is in micrometers. Real indices of refraction of air and liquid
water are given in Table 9.3 for several wavelengths.
Example 9.4
Find the angle of refraction in water if light of wavelength = 0.5 m enters
water from air at an incident angle
1
of 45
. Also nd the speed of light in air

and water at this wavelength.
SOLUTION
At = 0.5 m, n
air
= 1.000 279 and n
water
= 1.335 from Table 9.3. Thus, from
(9.19), the angle of refraction in water is
2
= 32
. From (9.20), the speeds

of light in air and water at = 0.5 m are 2.9971 10
8
m s
1
and 2.2456
10
8
m s
1
, respectively.
285
Blue
Red
Blue
Red
42

40
Visible
radiation
Figure 9.9 Geometry of a primary rainbow.
9.3.3 Dispersion
Refraction affects atmospheric optics in several ways. Rainbows appear when
incident light is refracted, dispersed, reected, then refracted again, as shown in
Fig. 9.9. As a beam of visible light enters a raindrop, all wavelengths bend toward
the surface normal due to refraction. Blue light bends the most and red light, the
least. The separation of white light into individual colors by selective refraction is
called dispersion (or dispersive refraction). When individual wavelengths hit the
back of a drop, they reect internally. When reected waves reach the front edge
of the drop, they leave the drop and refract away from the surface normal. Only
one wavelength from each raindrop impinges upon a viewers eye. Thus, a rain-
bowappears when individual waves frommany raindrops are seen simultaneously.
Red light appears on the top and blue (or violet) light appears on the bottom of
a primary rainbow (Fig. 9.9). The overall angle between the incident beam and
the beam reaching a viewers eye following interaction with rain is 40
for blue
light and 42
for red light (Fig. 9.9). A secondary rainbow can occur if a second
reection occurs inside each drop. The geometry of a rainbow is such that the Sun
must be at a viewers back for the viewer to see a rainbow.
9.3.4 Diffraction
Diffraction is a process by which the direction that a wave propagates changes
when the wave encounters an obstruction. Diffraction bends waves as they
pass by the edge of an obstruction, such as an aerosol particle, cloud drop, or
raindrop.
Diffraction can be explained in terms of Huygens principle, which states that
each point of an advancing wavefront may be considered the source of a new
series of secondary waves. A wavefront is a surface of constant phase in a waves
motion. If a stone is dropped in a tank of water, waves move out horizontally in
all directions, and wavefronts are seen as concentric circles around the stone. If a
point source emits waves in three dimensions, wavefronts are concentric spherical
286
9.3 Light processes
Primary
wavefronts
Source
Particle
Secondary
wavefronts
Diffracted rays
A
Figure 9.10 Diffraction around a particle.
Any point along a wavefront may be taken
as the source of a new series of secondary
waves. Rays emitted from point A appear to
cause waves fromthe original source to bend
around the particle.
surfaces. When a wavefront encounters an obstacle, such as in Fig. 9.10, waves
appear to bend (diffract) around the obstacle because a series of secondary con-
centric waves is emitted at the surface of the obstacle (and at other points along
the primary wavefronts). New waves do not appear in the backward direction,
because the intensity of the secondary wavelet depends on angle and is zero in the
backward direction, as Kirchoff demonstrated.
A corona is a set of prismatic colored rings around the Sun or Moon that arises
when sunlight or moonlight diffract around spherical liquid water drops in thin
clouds slightly blocking the Sun or Moon. Sometimes, the corona around the Moon
appears white with alternating bands of light and dark. These alternating bands are
caused by constructive and destructive interference, which arise due to the bend-
ing of some waves more than others during diffraction. Constructive interference
results when the crests or troughs of two waves meet, producing a band of bright
light. Destructive interference results when a crest meets a trough, producing a
band of darkness.
9.3.5 Particle scattering
Particle scattering is the combination of the effects of reection, refraction, and
diffraction. When a wave approaches a spherical particle, such as a cloud drop, it
can reect off the particle, diffract around the edge of the particle, or refract into
the particle. Once in the particle, the wave can be absorbed, transmit through the
particle and refract out, or reect internally one or more times and then refract
out. Figure 9.11 illustrates these processes, except for absorption, which is not a
scattering process. The processes that affect particle scattering the most are diffrac-
tion and double refraction, identied by rays C and B, respectively. Thus, parti-
cles scatter light primarily in the forward direction. They also scatter some light
to the side and in the backward direction. Backscattered light results primarily
287
A Sidescattering
B
C
D Sidescattering
E
Backscattering
Forward
scattering
Figure 9.11 Radiative scattering by a sphere. Ray A
is reected, B is refracted twice, C is diffracted, D is
refracted, internally reected twice, then refracted, and
E is refracted, reected once, then refracted. Rays A, B,
C, and D scatter in the forward or sideward direction. E
scatters in the backward direction.
Backscattering Forward scattering
Cloud drop
S
u
n
l
i
g
h
t
Figure 9.12 Forward scattering and backscattering by a cloud
drop. Forward scattering is due primarily to diffraction and dou-
ble refraction. Backscattering is due primarily to a single internal
reection.
from a single internal reection (ray E). Figure 9.12 shows the primary processes
by which a cloud drop might scatter radiation in the forward and backward
directions.
Whether one or more internal reections occurs within a particle depends on
the angle at which a wave within the particle strikes a surface. Suppose light in
Fig. 9.8 travels from water to air instead of from air to water. In the gure, total
internal reection within the water occurs only if
1
90
. From (9.19),
1
= 90
when
2.c
= sin
1
n
1
n
2
sin 90
(9.22)
which is the critical angle. When
2
-
2.c
, some light reects internally, but most
refracts out of the water to the air. When
2

2.c
, total internal reection occurs
within the water. Total internal reection can occur only when light originates from
the medium of lower index of refraction.
288
9.3 Light processes
Example 9.5
Find the critical angle of light at = 0.5 m in a liquid water drop suspended
in air.
SOLUTION
From Table 9.3, the indices of refraction of water and air are 1.335 and
1.000 279, respectively. Substituting these values into (9.22) gives
2.c
= 48.53.
Thus, total internal reection within the drop occurs when
2
48.53
.
9.3.6 Gas scattering
Gas scattering is the redirection of radiation by a gas molecule without a net transfer
of energy to the molecule. When a gas molecule scatters, incident radiation is
redirected symmetrically in the forward and backward direction and somewhat
off to the side. This differs from cloud drops and aerosol particles, which scatter
mostly in the forward direction. Visible radiation does not diffract readily around
gas molecules, because visible wavelengths are much larger than is the diameter of
a gas molecule.
Another difference between gas scattering and particle scattering is that gases
selectively scatter the shortest (blue) wavelengths of light whereas particles scatter
all wavelengths relatively equally. Selective scattering by gas molecules explains
why the Sun appears white at noon, yellow in the afternoon, and red at sunset and
why the sky is blue.
White sunlight that enters the Earths atmosphere travels a shorter distance
before reaching a viewers eye at noon than at any other time during the day, as
illustrated in Fig. 9.13. During white lights travel through the atmosphere, blue
wavelengths are preferentially scattered out of the direct beam by gas molecules,
but not enough blue light is scattered for a person looking at the Sun to notice
that the incident beam has changed its color from white. Thus, a person looking
at the Sun at noon often sees a white Sun. The blue light that scatters out of the
direct beam is scattered by gas molecules multiple times, and some of it eventually
enters the viewers eye when the viewer looks away fromthe Sun. As such, a viewer
looking away from the Sun sees a blue sky.
In the afternoon, light takes a longer path through the air than it does at noon;
thus, more blue and some green light is scattered out of the direct solar beam in
the afternoon than at noon. Although a single gas molecule is less likely to scatter
a green than a blue wavelength, the number of gas molecules along a viewers line
of sight is so large in the afternoon that the probability of green light scattering is
sizable. Nearly all red and some green are still transmitted to the viewers eye in
the afternoon, causing the Sun to appear yellow. In clean air, the Sun can remain
yellow until just before it reaches the horizon.
When the Sun reaches the horizon at sunset, sunlight traverses its longest distance
through the atmosphere, and all blue and green and some red wavelengths are
289
Blue
Red
Earth Atmosphere
Space
White
Noon
Yellow
Afternoon
Sunset /
Twilight
Red
Green
Blue
Blue
Green
Red
Figure 9.13 Colors of the Sun. At noon, the Sun
appears white because red, green, and some blue
light transmit to a viewers eye. In the afternoon,
sunlight traverses a longer path in the atmosphere,
removing more blue. At sunset, most green is
removed from the line of sight, leaving a red Sun.
After sunset, the sky appears red due to refraction
between space and the Earths atmosphere.
scattered out of the Suns direct beam. Only some direct red light transmits, and a
viewer sees a red Sun. Sunlight can be seen after sunset because sunlight refracts as
it enters Earths atmosphere, as illustrated in Fig. 9.13.
9.4 ABSORPTION AND SCATTERING BY
GASES AND PARTICLES
Whereas scattering redirects radiation, absorption removes radiation from an inci-
dent beam. In both cases, radiation in the beam is attenuated, reducing the quan-
tity of radiation transmitted. Aerosol particles, cloud drops, and gases scatter and
absorb as a function of wavelength. In polluted, cloud-free air, the main process
reducing visibility is aerosol particle scattering. Absorption of visible light is impor-
tant only when soot (black carbon plus organic matter) is present. Gas absorption
is important only when nitrogen dioxide concentrations are high. Gas scattering
always occurs but is important relative to other processes only in clean air. In this
section, gas absorption, gas scattering, particle absorption, and particle scattering
are discussed in some detail.
9.4.1 Gas absorption
Gases selectively absorb radiation as a function of wavelength. In this subsection,
gas absorption in the solar and infrared spectra are discussed, the gas absorption
extinction coefcient is dened, and the effect of gas absorption on visibility is
analyzed.
290
9.4 Absorption and scattering by gases and particles
Table 9.4 Wavelengths of absorption in the visible and UV spectra by several gases
Gas name Chemical formula Absorption wavelengths (m)
Visible/near-UV/far-UV absorbers
Ozone O
3
-0.35, 0.450.75
Nitrate radical NO
3
-0.67
Nitrogen dioxide NO
2
-0.71
Near-UV/far-UV absorbers
Nitrous acid HONO -0.4
Dinitrogen pentoxide N
2
O
5
-0.38
Formaldehyde HCHO -0.36
Hydrogen peroxide H
2
O
2
-0.35
Acetaldehyde CH
3
CHO -0.345
Peroxynitric acid HO
2
NO
2
-0.33
Nitric acid HNO
3
-0.33
Peroxyacetyl nitrate CH
3
CO
3
NO
2
-0.3
Far-UV absorbers
Molecular oxygen O
2
-0.245
Nitrous oxide N
2
O -0.24
CFC-11 CFCl
3
-0.23
CFC-12 CF
2
Cl
2
-0.23
Methyl chloride CH
3
Cl -0.22
Carbon dioxide CO
2
-0.21
Water vapor H
2
O -0.21
Molecular nitrogen N
2
-0.1
9.4.1.1 Gas absorption in the solar spectrum
Gas absorption in the solar spectrum is important for three reasons: (1) absorption
heats the air and prevents some solar radiation, including harmful UV radiation,
fromreaching the ground, (2) absorption often breaks gases into smaller molecules
or atoms during photolysis, driving atmospheric photochemistry, and (3) absorp-
tion affects visibility.
When a gas absorbs solar radiation, it converts electromagnetic energy to inter-
nal energy, increasing the thermal speed of the gas, thereby raising its temperature.
Table 9.4 shows several gases that absorb UV and visible radiation. The most
noticeable effects of gas absorption on air temperature are the stratospheric inver-
sion, which is caused entirely by absorption of UV light by ozone (O
3
), and the
thermospheric inversion, caused by absorption of UV light by molecular nitrogen
(N
2
2
) (Fig. 2.4).
In 1880, M. J. Chappuis found that ozone absorbs visible radiation of 0.45
0.75 m (now the Chappuis bands). In 1916, English physicists Alfred Fowler
(18681940) and Robert John Strutt (18751947, the son of Lord Rayleigh)
showed that ozone also weakly absorbs at 0.310.35 m(nowthe Huggins bands).
When ozone absorbs in the Huggins and Chappuis bands, it often photolyzes to
O
2
+ O(
3
P). In 1881, John Hartley suggested that ozone was present in the upper
atmosphere and hypothesized that the reason that the Earths surface received little
radiation shorter than 0.31 m was because ozone absorbed these wavelengths.
291
The absorption bands of ozone below 0.31 m are now the Hartley bands. When
ozone absorbs in the Hartley bands, it often photolyzes to O
2
+ O(
1
D). Ozone
absorption prevents nearly all UV wavelengths 0.2450.28 m and most wave-
lengths 0.280.32 m from reaching the troposphere, protecting the surface of the
Earth.
Molecular oxygen absorbs wavelengths -0.245m and molecular nitrogen
absorbs wavelengths -0.1m, together preventing nearly all solar wavelengths
-0.245m from reaching the troposphere. Oxygen absorption at wavelengths
-0.175m (an absorption wavelength region called the SchumannRunge sys-
tem) often results in its photolysis to O(
1
D) +O(
3
P). Absorption from 0.175 to
0.245m (the Herzberg continuum) often results in its photolysis to 2O(
3
P).
Although the gases in Table 9.4, aside from O
3
, O
2
, and N
2
, absorb UV radia-
tion, their mixing ratios are too low to reduce it signicantly. For instance, strato-
spheric mixing ratios of water vapor (0 to 6 ppmv) are much lower than are those
of O
2
, which absorbs many of the same wavelengths as does water vapor. Thus,
water vapor has little effect on UVattenuation in the stratosphere. Similarly, strato-
spheric mixing ratios of carbon dioxide (370 ppmv) are much lower than are those
of O
2
or N
2
, both of which absorb the same wavelengths as does carbon dioxide.
Although water vapor mixing ratios are much higher in the troposphere than in
the stratosphere, wavelengths that water absorbs do not reach the troposphere.
Wavelengths reaching the troposphere are longer than 0.28m.
Of the gases listed in Table 9.4, only ozone, nitrogen dioxide (NO
2
), and the
nitrate radical (NO
3
) absorb in the visible spectrum. The rest absorb in the ultra-
violet spectrum. Absorption by ozone is weak in the visible spectrum, and con-
centrations of the nitrate radical are relatively low, except at night, when sun-
light is absent. Thus, in polluted air, nitrogen dioxide is the only gas that absorbs
enough radiation to affect visibility. Nitrogen dioxide appears yellow, brown, or
red because it absorbs blue light preferentially and green light to a lesser extent.
9.4.1.2 Gas absorption in the infrared spectrum: the greenhouse effect
Gas absorption in the solar-infrared and thermal-infrared affects the Earths energy
balance. Such absorption does not affect photolysis signicantly or visibility at all.
Figure 9.14 shows the fraction of radiation incident at the top of Earths atmo-
sphere that is transmitted to the surface, versus wavelength, in the presence of
individual gases and the combination of multiple gases. The gases listed in the
gure (H
2
O, CO
2
, CH
4
, CO, O
3
, O
2
, N
2
O, CH
3
Cl, CFCl
3
, CF
2
Cl
2
, and CCl
4
)
are greenhouse gases, which are gases that are relatively transparent to incom-
ing visible radiation but opaque to selective wavelengths of outgoing infrared
radiation.
The natural greenhouse effect is the warming of the Earths lower atmosphere
due to natural greenhouse gases. Greenhouse gases cause a net warming of the
Earths atmosphere like a glass house causes a net warming of its interior. Because
most incoming solar radiation can penetrate a glass house but a portion of outgoing
thermal-infrared radiation cannot, air inside a glass house warms during the day
292
0
0.2
0.4
0.6
0.8
1
1 10 100 1000
F
r
a
c
t
i
o
n

t
r
a
n
s
m
i
t
t
e
d
Wavelength (m)
H
2
O
(a)
0
0.2
0.4
0.6
0.8
1
1 10 100 1000
F
r
a
c
t
i
o
n

t
r
a
n
s
m
i
t
t
e
d
Wavelength (m)
CO
2
(b)
0
0.2
0.4
0.6
0.8
1
1 10 100 1000
F
r
a
c
t
i
o
n

t
r
a
n
s
m
i
t
t
e
d
Wavelength (m)
O
3
(c)
0.7
0.75
0.8
0.85
0.9
0.95
1
1 10 100 1000
F
r
a
c
t
i
o
n

t
r
a
n
s
m
i
t
t
e
d
Wavelength (m)
CH
4
(d)
0
0.2
0.4
0.6
0.8
1
1 10 100 1000
F
r
a
c
t
i
o
n

t
r
a
n
s
m
i
t
t
e
d
Wavelength (m)
Eleven gases
(e)
Figure 9.14 Comparison of the wavelength-dependent fraction of transmitted incident radia-
tion, calculated from line-by-line data () (Rothman et al. 2003) with that calculated from a
model (-----) (Jacobson 2005a) when absorption by (a) H
2
O, (b) CO
2
, (c) O
3
, (d) CH
4
, and
(e) 11 absorbing gases together, is considered. Mass pathlengths (g cm
2
) of the gases, dened in
(9.26), were as follows: H
2
O: 4.7275; CO
2
: 0.58095; O
3
: 0.00064558; CH
4
: 0.00092151; O
2
:
239.106; N
2
O: 0.000461839; CO: 8.511110
5
; CH
3
Cl: 9.100641 10
7
; CCl
4
: 4.69957
10
7
; CFCl
3
: 8.5305910
7
; CF
2
Cl
2
: 1.39333 10
6
. These pathlengths represent path-
lengths through the entire atmosphere, but the calculations were performed at a constant
temperature (270 K) and pressure (322.15 hPa).
so long as mass (such as plant mass) is present to absorb solar and reemit thermal-
infrared radiation. The surface of the Earth, like plants, absorbs solar and reemits
thermal-infrared radiation. Greenhouse gases, like glass, are transparent to most
solar radiation but absorb a portion of thermal-infrared.
The natural greenhouse effect is responsible for about 33 K of the Earths aver-
age near-surface air temperature of 288 K. Without the natural greenhouse effect,
Earths average near-surface temperature would be about 255 K, which is too
cold to sustain most life. Thus, the presence of natural greenhouse gases is ben-
ecial. Human emission of greenhouse gases has increased the concentrations of
293
x
0
x
I
0,
I
dx
Figure 9.15 Attenuation of
incident radiance I
0
due to
absorption in a column of gas.
such gases, causing global warming. Global warming is the increase in the Earths
temperature above that from the natural greenhouse effect due to the addition of
anthropogenically emitted greenhouse gases and particulate black carbon to the
air.
The most important greenhouse gas is water vapor, which accounts for approx-
imately 90 percent of the 33-K temperature increase due to natural greenhouse
warming. Figure 9.14(a) shows the infrared wavelengths at which water vapor
absorbs (and transmits). Carbon dioxide is the second most important and abun-
dant natural greenhouse gas. Its transmission spectrum is shown in Fig. 9.14(b).
Both water vapor and carbon dioxide are transparent to most visible radiation.
Most infrared absorption by water vapor and carbon dioxide occurs outside the
wavelength region, 812m, called the atmospheric window. The atmospheric
window can be seen in Fig. 9.14(e), which shows the transmission through all
important infrared absorbers in the atmosphere. In the atmospheric window, gases
are relatively transparent to thermal-infrared radiation, allowing the radiation to
radiate to higher altitudes. Ozone (Fig. 9.14(c)), several chlorouorocarbons, and
methyl chloride absorb radiation within the atmospheric window. Nitrous oxide
and methane (Fig. 9.4(d)) absorb at the edges of the window. Increases in concen-
tration of these gases enhance global warming by strengthening thermal-infrared
absorption at wavelengths that usually radiate to higher altitudes.
9.4.1.3 Gas absorption extinction coefcient
The gas absorption extinction coefcient,
a.g.
(cm
1
, m
1
, or km
1
) quanties
the extent of absorption by a gas. An extinction coefcient describes the loss of
electromagnetic radiation due to a specic process, per unit distance, and may be
determined as the product of an effective cross section and a number concentration.
In the case of gas absorption, the extinction coefcient through a uniformly mixed
gas q is
a.g.q..T
= N
q
b
a.g.q..T
, where N
q
is the number concentration of the gas
(molec. cm
3
) and b
a.g.q..T
is the absorption cross section of the gas (cm
2
) at
wavelength and temperature T. The absorption cross section of a gas is an effective
cross section that results in radiance reduction by absorption.
Suppose incident radiance I
0.
travels a distance dx through the uniformly mixed
absorbing gas shown in Fig. 9.15. The reduction in radiance with distance through
294
the gas is
dI
dx
= N
q
b
a.g.q..T
I
=
a.g.q..T
I
(9.23)
Integrating (9.23) from I
0.
to I
and x
0
to x gives
I
= I
0.
e
N
q
b
a.g.q..T
(xx
0
)
= I
0.
e
a.g.q..T
(xx
0
)
(9.24)
Equation (9.24) states that incident radiation decreases from I
0.
to I
over distance
dx by gas absorption. The absorption extinction coefcient quanties the rate of
absorption per unit distance.
Absorption cross-section data can be extracted mathematically from (9.24) if
the attenuation of radiance from I
0.
to I
through a column of gas of molec-

ular pathlength N
q
(x x
0
) (molec. cm
2
) is measured experimentally. DeMore
et al. (1997) and Atkinson et al. (1997) provide absorption cross-section data as a
function of wavelength and temperature calculated in this manner.
For radiative transfer calculations, the extinction coefcient due to gas absorp-
tion, summed over all absorbing gases, is required. This quantity is
a.g.
=
N
ag
q=1
N
q
b
a.g.q..T
=
N
ag
q=1
a.g.q..T
(9.25)
where N
ag
is the number of absorbing gases, and the subscript T was dropped in
a.g.
. In the infrared spectrum, the gas absorption extinction coefcient may be
written in terms of a mass absorption coefcient k
a.g.q.
(cm
2
g
1
),
a.g.
=
N
ag
q=1
q
k
a.g.q.
=
N
ag
q=1
N
q
m
q
A
k
a.g.q.
=
N
ag
q=1
u
q
(x x
0
)
k
a.g.q.
(9.26)
where
q
= N
q
m
q
Ais the mass density (g cm
3
), N
q
is the number concentration
(molec. cm
3
), m
q
is the molecular weight (g mol
1
), u
q
=
q
(x x
0
) is the mass
pathlength (g cm
2
) of gas q in air, and A is Avogadros number (molec. mol
1
).
Mass absorption coefcients are cross sections, averaged over a wavelength incre-
ment, per unit mass of absorbing gas.
Mass absorption coefcients can be obtained fromhigh-resolution spectral data,
such as fromthe HITRANdatabase (Rothmanet al. 2003). This particular database
provides absorption parameters for over one million narrow absorption line distri-
butions in the solar- and thermal-infrared spectra. Line distributions are functions
describing the intensity of gas absorption over a narrow region in wavelength
space. The width and peak intensity of absorption of each line distribution of a gas
varies with temperature and pressure. Line distributions may overlap one another.
Line-distribution data are commonly referred to as line-by-line data.
The shape of a line distribution is often modeled with the Lorentz prole (Lorentz
1906; Rothman et al. 2003), examples of which are shown in Fig. 9.16. With this
295
10
3
10
2
10
1
10
0
10
1
10
2
6.619 6.6205 6.622 6.6235 6.625
A
b
s
o
r
p
t
i
o
n

c
o
e
f
f
i
c
i
e
n
t

(
c
m
2
/
g
)
Wavelength (m)
Figure 9.16 Absorption coefcient of
water vapor versus wavelength over
one line distribution with a Lorentzian
shape prole at two pressures,
322.15 hPa, ----22.57 hPa and at a
temperature of 270 K. The curve was
calculated with (9.27) with data from
Rothman et al. (2003).
prole, the absorption coefcient (cm
2
g
1
) of gas q at wavelength is calculated
with
k
a.g.q.
=
A
m
q
1
S
q
(T)
q
( p
a
. T)
q
( p
a
. T)
2
+[
(
q
q
( p
ref
) p
a
)]
2
(9.27)
where A is Avogadros number (molec. mol
1
), m
q
is the molecular weight of gas
q (g mol
1
), S
q
(T) is the temperature-dependent intensity of the line distribution
(cm
1
(molec.
1
cm
2
)),
q
( p
a
. T) is the pressure-broadened halfwidth of the line
distribution (cm
1
),
is the wavenumber (cm

1
) at wavelength ,
q
is the central
wavenumber (cm
1
) of the line distribution before an air-broadened pressure shift,
q
( p
ref
) is the air-broadened pressure shift (cm
1
atm
1
) of wavenumber
q
at pres-
sure p
ref
= 1 atm, p
a
is current atmospheric pressure (atm), and
q
q
( p
ref
) p
a
is
the central wavenumber of the distribution after adjustment for pressure (cm
1
).
The pressure-broadened halfwidth is
q
( p
a
. T) =
T
ref
T
n
q
[
air.q
( p
ref
. T
ref
)( p
a
p
q
) +
self.q
( p
ref
. T
ref
) p
q
] (9.28)
where T
ref
= 296 K, T is current temperature (K), n
q
is a coefcient of tempera-
ture dependence,
air.q
( p
ref
. T
ref
) is the air-broadened halfwidth (cm
1
atm
1
) at T
ref
and p
ref
, p
q
is the partial pressure of gas q (atm), and
self.q
is the self-broadened
halfwidth (cm
1
atm
1
) of gas q at T
ref
and p
ref
. The parameter S
q
(T) is calculated as
a function of temperature from S
q
(T
ref
), which, along with n
q
,
air.q
( p
ref
. T
ref
), and
296
self.q
( p
ref
. T
ref
), are available from the HITRAN database (Rothman et al. 2003).
Figure 9.16 shows absorption coefcient versus wavelength, calculated from(9.27),
from one line distribution at two pressures. At high pressure, the line distribution
is broader than at low pressure.
Greenhouse gases, such as H
2
O and CO
2
, absorb in several tens of thousands of
line distributions, such as those shown in Fig. 9.14, across the solar- and thermal-
infrared spectra. Because atmospheric models discretize the wavelength spectrum
into only tens to hundreds of wavelength intervals, a method is needed to gather
line-distribution information into each model interval for practical use.
Merely averaging line-by-line absorption coefcients from several line distribu-
tions over each model wavelength interval, though, overestimates absorption. This
can be demonstrated as follows. The monochromatic (single-wavelength) trans-
mission of radiation through an absorbing gas is the fraction of incident radiance
that penetrates through the gas, and is dened as
T
a.g.q.
(u
q
) =
I
I
0
= e
u
q
k
a.g.q.
= e
a.g.q.
(xx
0
)
(9.29)
where u
q
=
q
(x x
0
) is the mass pathlength of the gas (g cm
2
), and
a.g.q.
=
q
k
a.g.q.
. Suppose a gas has two nonoverlapping absorption line distributions,
equally spaced within a model wavenumber interval, to +L , but of different
strength, k
x
and k
y
(where the subscripts a, g, and q were omitted). The exact
transmission through pathlength u of the gas within the interval is
T
. +L
(u) = 0.5(e
k
x
u
+e
k
y
u
) (9.30)
If the absorption coefcients are rst averaged, the resulting transmission is
T
. +L
(u) = e
0.5(k
x
+k
y
)u
(9.31)
which is always smaller (more absorbing) than the exact solution, (9.30), except
when k
x
= k
y
. Thus, averaging multiple absorption coefcients overestimates
absorption.
Amethod of addressing the overabsorption problemis the k-distribution method
(e.g., Ambartzumiam 1936; Kondratyev 1969; Yamamoto et al. 1970; Arking and
Grossman 1972; Lacis and Hansen 1974; Liou 2002). With this method, absorption
line distributions in each model wavenumber interval are grouped into typical
probability intervals according to absorption coefcient strength. For example,
the strongest absorption coefcients are grouped together, the next strongest are
grouped together, and so on. The absorption coefcients within each probability
interval are then averaged, and the transmission in the wavenumber interval as
a whole is calculated by integrating the transmission through each probability
interval with the rightmost expression in
T
. +L
(u) =
1
L
+L

e
k

u
d

0
e
k
u
f

(k
)dk
(9.32)
297
where u is the pathlength of the gas through the layer of air considered, k

is the
monochromatic (single-wavenumber) absorption coefcient, and k
is the absorp-
tion coefcient within a probability interval of normalized probability f

(k
) (or
differential probability f

(k
)dk
) of occurring. The integral of the normalized prob-

ability, over all possible absorption coefcient strengths, is unity:

0
f

(k
)dk
= 1 (9.33)
The middle expression in (9.32) is the exact solution, but requires integration over
tens of thousands of individual spectral subintervals. The rightmost integral is
generally approximated as a summation over 416 probability intervals. With the
k-distribution method, probability intervals of a gas differ for each layer of the
atmosphere.
An extension of the original method is the correlated k-distribution method (e.g.,
Lacis et al. 1979; Goody et al. 1989; West et al. 1990; Lacis and Oinas 1991; Fu
and Liou 1992; Stam et al. 2000; Liou 2002). With this method, the wavenumber
order of absorption coefcients is rst rearranged by strength into a cumulative
wavenumber frequency distribution,
g

(k) =
k
0
f

(k
)dk
(9.34)
where g

(k) varies between 0 and 1 and is a monotonically increasing function of
the absorption coefcient k. The cumulative distribution is then divided up into
probability intervals. The probability intervals are then assumed to be the same
(correlated) at each altitude for a given gas. Finally, radiative transfer is calculated
through the atmosphere for each probability interval, and the resulting transmission
is weighted among all probability intervals to obtain a net transmission.
When multiple gases are considered, the complexity deepens because the relative
strength of absorption of one gas at a given wavelength depends on the quantity
of radiation removed by all other gases at the same wavelength. Methods of treat-
ing absorption among multiple gases are given in West et al. (1990), Lacis and
Oinas (1991), and Jacobson (2005a). With the last method, a major absorber is
selected for each model wavelength interval. Absorption coefcients calculated in
each model subinterval for the major absorber are then grouped by strength in each
probability interval. Finally, model subintervals containing absorption coefcients
for other gases are grouped into the same probability intervals as those of the major
gas according to wavenumber. In other words, if a subinterval of the major gas of
a given wavenumber is placed in a given probability interval, the subintervals of
all other gases at that same wavenumber are placed in the same probability inter-
val. This technique, called the multiple-absorber correlated-k-distribution method,
ensures exact correlation of absorption wavelengths (wavenumbers) within a
probability interval for all gases through all layers of the atmosphere when mul-
tiple gases are considered. The technique is physical since all gases in reality are
correlated in wavelength space. Figure 9.14 compares modeled with line-by-line
298
Table 9.5 Extinction coefcients (
a,g
) and meteorological ranges (x
a,g
) due to NO
2
absorption at selected wavelength intervals ( 0.005m) and concentrations
0.01 ppmv NO
2
0.25 ppmv NO
2
b
a,g
x
a,g

a,g
x
a,g
x
s,g
(m) (10
19
cm
2
) (10
8
cm
1
) (km) (10
8
cm
1
) (km) (km)
0.42 5.39 13.2 296 330 11.8 112
0.45 4.65 11.4 343 285 13.7 148
0.50 2.48 6.10 641 153 25.6 227
0.55 0.999 2.46 1590 61.5 63.6 334
0.60 0.292 0.72 5430 18.0 217 481
0.65 0.121 0.30 13000 7.5 520 664
Absorption-cross-section data (b) for NO
2
are from Schneider et al. (1987). Also shown is the
meteorological range due to Rayleigh scattering only (x
s,g
). T = 298 K and p
a
= 1 atm.
absorption coefcients for individual gases and multiple gases together from this
technique.
Once an absorption coefcient of an individual gas is calculated, it is applied in
(9.26) to determine the extinction coefcient of the gas at the current mass density
of the gas.
9.4.1.4 Gas absorption effects on visibility
In the solar spectrum, gas absorption, primarily by nitrogen dioxide, affects
visibility. The furthest distance a typical eye can see at a given wavelength
may be estimated with the Koschmieder equation, x
= 3.912
ext.
, derived in
Section 9.5. In the equation, x
is the meteorological range and

ext.
is the total
extinction coefcient, which includes extinction due to gas absorption, gas scatter-
ing, particle absorption, and particle scattering.
Table 9.5 gives extinction coefcients and meteorological ranges due to NO
2
absorption, alone. It shows that NO
2
absorbs more strongly at shorter (blue) than
at longer (green or red) wavelengths. At low concentrations (0.01 ppmv), the effect
of NO
2
absorption on visibility is less than that of gas scattering at all wavelengths.
At typical polluted-air concentrations (0.10.25 ppmv), NO
2
reduces visibility sig-
nicantly for wavelengths -0.50 mand moderately for wavelengths 0.5 0.6 m.
Most effects of NO
2
on visibility are limited to times when its concentration
peaks. Results froma project studying Denvers brown cloud, for example, showed
that NO
2
accounted for about 6.6 percent of total extinction averaged over all sam-
pling periods, and 37 percent of extinction during periods of maximum NO
2
con-
centration. Scattering and absorption by particles caused most remaining extinction
(Groblicki et al. 1981).
Figure 9.17 shows extinction coefcients due to NO
2
and ozone absorption
at different mixing ratios. The gure indicates that NO
2
affects extinction (and
therefore radiative transfer and visibility) only at high mixing ratios and at wave-
lengths below about 0.5m. In polluted air, such as in Los Angeles, NO
2
mixing
ratios typically range from 0.01 to 0.1 ppmv and peak near 0.15 ppmv during the
299
10
6
10
5
10
4
10
3
10
2
10
1
10
0
10
1
0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7
E
x
t
i
n
c
t
i
o
n

c
o
e
f
f
i
c
i
e
n
t

(
k
m
1
)
Wavelength (m)
NO
2
(g) (0.25 ppmv)
O
3
(g) (0.25 ppmv)
NO
2
(g) (0.01 ppmv)
O
3
(g) (0.01 ppmv)
Figure 9.17 Extinction coefcients due to NO
2
and O
3
absorption when T = 298 K and
p
a
= 1013 hPa. Cross-section data for NO
2
are from DeMore et al. (1997) and interpolated
fromSchneider et al. (1987). Data for ozone are
from Atkinson et al. (1997).
morning. A typical value is 0.05 ppmv, which results in an extinction coefcient
of about 0.07 km
1
at 0.4 m and 0.01 km
1
at 0.55 m. The visibility (meteo-
rological range) is about 390 km with this latter value. Ozone has a larger effect
on extinction than does nitrogen dioxide at wavelengths below about 0.32 m.
Ozone mixing ratios in polluted air usually peak between 0.05 and 0.25 ppmv.
Nevertheless, the cumulative effect of ozone, nitrogen dioxide, and other gases on
extinction is small in comparison with the effects of scattering and absorption by
particles.
9.4.2 Gas scattering
The only gas-scattering process in the atmosphere is Rayleigh scattering, which
is the scattering of radiation by gas molecules (primarily N
2
and O
2
). Rayleigh
scattering gives the sky its blue color. It causes dark objects a few kilometers away
to appear behind a blue haze of scattered light, and bright objects more than
30 km away to appear reddened (Waggoner et al. 1981).
A Rayleigh scatterer has molecular radius r much smaller than the wavelength
of interest (2r 1). The radius of a gas molecule is sufciently small in relation
to visible-light wavelengths to meet this criterion. Since oxygen and nitrogen are
the most abundant gases in the atmosphere, they are the most abundant Rayleigh
scatterers.
The extinction coefcient (cm
1
) due to Rayleigh scattering is
s.g.
= N
a
b
s.g.
(9.35)
where N
a
is the number concentration of air molecules (molec. cm
3
) at a given
altitude, and b
s.g.
is the scattering cross section of a typical air molecule. Since
the number concentration of air molecules decreases exponentially with increasing
300
altitude, (9.35) suggests that extinction due to Rayleigh scattering also decreases
exponentially with increasing altitude.
The scattering cross section can be estimated with
b
s.g.
=
8
3
n
2
a.
1
2
3
4
N
2
a.0
f (
)
32
3
(n
a.
1)
2
3
4
N
2
a.0
f (
) (9.36)
where n
a.
is the real index of refraction of air, obtained from (9.21), N
a.0
=
2.55 10
19
cm
3
is the number concentration of air molecules at standard temper-
ature and pressure (288 K and 1 atm), and f (
) is the anisotropic correction term.

This term accounts for deviations of Rayleigh scattering from perfect isotropic
scattering and can be estimated with
f (
) =
6 +3
6 7
1.05 (9.37)
where
0.0279 (Young 1980). f (
) is a function of wavelength and varies from

1.08 at = 0.2 m to 1.047 at = 1.0 m (Liou 2002). The second expression
in (9.36) was found by assuming n
2
a.
1 2(n
a.
1), which is possible because
n
a.
1.
Equation (9.36) suggests that the Rayleigh scattering cross section varies
inversely with the fourth power of wavelength. In other words, gases scatter short
wavelengths much more effectively than they do long wavelengths. Since the mete-
orological range is inversely proportional to the extinction coefcient, the mete-
orological range should then increase with increasing wavelength when only gas
scattering is considered. Table 9.5 conrms this supposition. It shows that, at short
visible wavelengths (e.g., 0.42 m), the meteorological range due to Rayleigh scat-
tering is much smaller than that at long wavelengths (e.g., 0.65 m).
Example 9.6
For = 0.5 m, p
a
= 1 atm (sea level), and T = 288 K,
s.g.
= 1.72
10
7
cm
1
from (9.35). This extinction coefcient results in a meteorologi-
cal range of x = 3.912
s.g.
= 227 km.
For = 0.55 m, the extinction coefcient decreases to
s.g.
= 1.17
10
7
cm
1
and the meteorological range increases to x = 334 km. Waggoner
et al. (1981) reported a total extinction at Bryce Canyon, Utah, corresponding
to a meteorological range of within a few percent of 400 km for a wavelength
of 0.55 m.
9.4.3 Particle absorption and scattering
Particle scattering is the most important solar radiation attenuation process in
polluted air, followed, in order, by particulate absorption, gas absorption, and
gas scattering. All particle components scatter solar and infrared radiation and
absorb infrared radiation, but few absorb solar radiation. For several species, the
301
ability to absorb solar radiation increases from the visible to ultraviolet spectra.
Below, important particle absorbers of radiation are discussed. Subsequently, meth-
ods of calculating particle scattering and absorption in an atmospheric model are
described.
9.4.3.1 Important particle absorbers
The strongest particle absorber of solar radiation is black carbon, the main com-
ponent of soot. Other absorbers of solar radiation include hematite (Fe
2
O
3
) and
aluminum oxide (alumina, Al
2
O
3
). Hematite is found in soil-dust particles. Some
forms of aluminum oxide are found in soil-dust particles and others are found in
combustion particles.
Particle absorbers in the UV spectrum include black carbon, hematite, alu-
minum oxide, and certain organic compounds. The organics absorb UV radiation
but less visible radiation. The strongest near-UVabsorbing organics include cer-
tain nitrated aromatics, polycyclic aromatic hydrocarbons (PAHs), benzaldehydes,
benzoic acids, aromatic polycarboxylic acids, and phenols (Jacobson 1999c). The
strong absorptivity of nitrated aromatics occurs because substitution of a nitrate
group onto a benzene ring shifts the peak absorption 0.057 m toward a longer
wavelength. The strong absorptivity of benzaldehydes and benzoic acids results
because the addition of an aldehyde or an acid group to a benzene ring shifts the
peak absorption 0.046 or 0.0255 m, respectively, toward a longer wavelength
(Dean 1992). Inorganic particle components with a near-UV absorption peak
include the nitrate ion (0.302 m), ammonium nitrate (0.308 m), and sodium
nitrate (0.297 m) (Cleaver et al. 1963; Sommer 1989). Although near-UV radi-
ation (0.250.38 m) makes up only 5 percent of total solar radiation, near-UV
wavelengths are responsible for most gas photolysis.
Most particle components are weak absorbers of visible and UV radiation.
Silicon dioxide (SiO
2
), which is the white, colorless, crystalline compound found in
quartz, sand, and other minerals, is an example. Sodium chloride (NaCl), ammo-
nium sulfate ((NH
4
)
2
SO
4
), and sulfuric acid (H
2
SO
4
) are also weak absorbers of
visible and UV radiation. Soil-dust particles, which contain SiO
2
, Al
2
O
3
, Fe
2
O
3
,
CaCO
3
, MgCO
3
, and other compounds, absorb both UV and visible radiation,
with stronger absorption in the UV (e.g., Gillette et al. 1993; Sokolik et al. 1993).
During heavy dust storms, particularly over desert regions, concentrations of soil
dust can decrease visibility to a few meters or tens of meters.
9.4.3.2 Particle refractive indices
Figure 9.18 shows a possible path of radiant energy through a particle. If scattering
is ignored, the attenuation, due to absorption, of radiant energy as it propagates
through the particle is approximately
dI
dx
=
4
I (9.38)
302
Table 9.6 Real and imaginary indices of refraction for some substances
at = 0.51 and 10.0m
0.5 m 10 m
Real (n
) Imaginary (
) Real (n
) Imaginary (
)
H
2
O(aq)
a
1.34 1.010
9
1.22 0.05
Soot(s)
b
1.82 0.74 2.40 1.0
Organic C(s)
b
1.45 0.001 1.77 0.12
H
2
SO
4
(aq)
b
1.43 1.010
8
1.89 0.46
(NH
4
)
2
SO
4
(s)
b
1.52 0.0005 2.15 0.02
NaCl(s)
b
1.45 0.00015 1.53 0.033
a
Hale and Querry (1973),
b
Krekov (1993).
x
0
x
I
0
I
dx
Figure 9.18 Atten-
uation of incident
radiance I
0
due
to absorption in a
particle.
where is the imaginary index of refraction, which quanties the extent to which
a substance absorbs radiation. The term 4 is an absorption extinction coef-
cient for a single particle. Although is a function of wavelength, the wave-
length subscript in (9.38) was dropped. Integrating (9.38) from I
0
to I and x
0
to x
gives
I = I
0
e
4(xx
0
)
(9.39)
The optical properties of a particle component can generally be derived from
a combination of its real and imaginary refractive indices. These parameters give
information about the particles ability to refract and absorb, respectively. Both
parameters are embodied in the complex refractive index,
m
= n
(9.40)
Table 9.6 gives the real (n
) and imaginary (
) indices of refraction for some

substances at =0.50 and 10 m. Soot has the largest imaginary index of refraction
in the visible spectrum among the substances shown.
303
Table 9.7 Light transmission through soot and liquid water
particles at = 0.50m
Transmission (II
0
)
Particle diameter (m) Soot ( = 0.74) Water ( = 10
9
)
0.1 0.16 0.999999997
1.0 8.0 10
9
0.99999997
10.0 0 0.9999997
0
0.1
0.2
0.3
0.4
0.5
0.2 0.3 0.4 0.5 0.6 0.7
I
m
a
g
i
n
a
r
y

i
n
d
e
x

o
f

r
e
f
r
a
c
t
i
o
n
Wavelength (m)
Figure 9.19 Imaginary index of
refraction of liquid nitrobenzene
versus wavelength. Data from Foster
(1992).
Figure 9.19 shows the imaginary index of refraction of liquid nitrobenzene as
a function of wavelength. Although liquid nitrobenzene has a low concentration
in atmospheric particles, the shape of its index-of-refraction curve is important
because it suggests that some organic particle components with peak absorption
wavelengths below 0.3 m absorb moderately between 0.3 and 0.4 m as well,
thereby affecting photolysis. Most organic atmospheric particles have peak absorp-
tion wavelengths below 0.3 m.
Table 9.7 shows the extent of transmission of light through soot and liquid water
particles at = 0.50 m. These calculations were obtained from (9.39). The trans-
mission (II
0
) is the fraction of incident radiation passing through a substance. The
table shows that a 0.1-m-diameter soot particle transmits 16 percent of incident
0.50-m radiation, and a 0.1-m water drop transmits nearly 100 percent of the
radiation if only absorption is considered. A 10-m liquid water drop absorbs a
negligible amount of incident 0.50-m radiation. Thus, soot is a strong absorber
of visible radiation, and liquid water is not.
9.4.3.3 Absorption and scattering efciencies
In a model, absorption and scattering extinction coefcients must be consid-
ered for both aerosol particles and hydrometeor particles. Aerosol absorption,
aerosol scattering, cloud absorption, and cloud scattering extinction coefcients are
304
calculated with
a.a.
=
N
B
i =1
n
i
b
a.a.i.

s.a.
=
N
B
i =1
n
i
b
s.a.i.
(9.41)
a.h.
=
N
B
i =1
n
i
b
a.h.i.

s.h.
=
N
B
i =1
n
i
b
s.h.i.
respectively, where n
i
is the number concentration of aerosol particles or hydrom-
eteor particles of a given size (particles cm
3
), N
B
is the number of size bins in
either case, b
a.a.i.
and b
a.h.i.
are the effective absorption cross sections of a single
aerosol particle and hydrometeor particle, respectively, and b
s.a.i.
and b
s.h.i.
are
the effective scattering cross sections of a single aerosol particle and hydrometeor
particle, respectively.
For spherical aerosol particles, the effective absorption and scattering cross sec-
tions are
b
a.a.i.
= r
2
i
Q
a.i.
b
s.a.i.
= r
2
i
Q
s.i.
(9.42)
respectively, where r
2
i
is the actual aerosol cross section (cm
2
), Q
a.i.
(dimen-
sionless) is the single-particle absorption efciency, and Q
s.i.
is the single-particle
scattering efciency. The absorption and scattering efciencies are functions of
the complex index of refraction m
, which depends on wavelength and particle

composition, and of the size parameter
i.
=
2r
i
(9.43)
which depends on particle size and wavelength. Absorption cross sections for
hydrometeor particles are calculated as in (9.42), but with single-particle absorp-
tion and scattering efciencies for hydrometeor particles rather than for aerosol
particles.
The single-particle scattering efciency is the ratio of the effective scattering
cross section of a particle to its actual cross section. The scattering efciency can
exceed unity, since a portion of the radiation diffracting around a particle can be
intercepted and scattered by the particle. Scattering efciencies above unity account
for this additional scattering. The greater the absorption by a particle, the lesser the
scattering efciency, since absorption hinders scattering. The scattering efciency
is highest when the wavelength of light is close to the particle radius.
The single-particle absorption efciency is the ratio of the effective absorption
cross section of a particle to its actual cross section. The absorption efciency can
exceed unity, since a portion of the radiation diffracting around a particle can
be intercepted and absorbed by the particle. Absorption efciencies above unity
account for this additional absorption. The larger the imaginary refractive index
of a particle, the greater its absorption efciency. The real refractive index affects
305
the angle that radiation bends upon entering a particle, affecting the distance it
travels and its cumulative absorption within the particle.
Single-particle absorption and scattering efciencies vary with particle size and
radiation wavelength. When a particles diameter is much smaller than the wave-
length of light (d
i
- 0.03, or
i.
- 0.1), the particle is in the Rayleigh regime
and is called a Tyndall absorber or scatterer. Such particles have an absorption
efciency of
Q
a.i.
= 4
2r
i
Im
m
2
1
m
2
+2
2r
i
24n
n
2
+
2
2
+4
n
2
+1
(9.44)
As
0, (9.44) approaches
Q
a.i.
=
2r
i
24n
n
2
+2
(9.45)
which is a linear function of
. When r
i
, the absorption efciency is small.
Thus, small particles are relatively inefcient absorbers of radiation.
The Tyndall scattering efciency is
Q
s.i.
=
8
3
2r
i
m
2
1
m
2
+2
2
(9.46)
Equations (9.46) and (9.44) show that the scattering and absorption efciencies of
small particles are proportional to (r
i
)
4
and r
i
, respectively. Thus, at a suf-
ciently short wavelength, the absorption efciency of an absorbing particle (large )
is much greater than is the scattering efciency of the same particle. Figures 9.20
and 9.21 show the single-particle absorption and scattering efciencies, including
Tyndall efciencies, of soot, a strong absorber, and of liquid water, a weak absorber.
Example 9.7
At = 0.5 m and r
i
= 0.01 m, nd the single-particle scattering and absorp-
tion efciencies of liquid water.
SOLUTION
From Table 9.6, n
= 1.34 and
= 1.0 10
9
at = 0.5 m. From (9.46) and
(9.45), respectively, Q
s.i.
= 2.92 10
5
and Q
a.i.
= 2.8 10
10
. Since liquid
water is nonabsorbing, its single-particle absorption efciency is expected to
be small in comparison with its single-particle scattering efciency.
306
0
0.5
1
1.5
2
0
0.5
1
1.5
2
0.01 0.1 1 10 100 1000
Particle diameter (m)
Q
a
Q
s
Q
f
Mie regime Geometric
regime
R
a
y
l
e
i
g
h

r
e
g
i
m
e
Figure 9.20 Single-particle absorption (Q
a
),
total scattering (Q
s
), and forward scattering
(Q
f
) efciencies of soot particles (which con-
tain mostly black carbon) of different sizes at
=0.50 m (n
=1.94,
=0.66). The ef-

ciency for soot at any other wavelength
1
in
the visible or UV spectrumand diameter d
1
is
the efciency in the gure located at diameter
d = 0.5md
1
1
.
0
1
2
3
4
5
10
9
10
7
10
5
10
3
10
1
10
1
10
3
0.01 0.1 1 10 100 1000
Q
a
Q
s
Q
f
Mie regime
Geometric
regime
R
a
y
l
e
i
g
h

r
e
g
i
m
e
Figure 9.21 Single-particle absorption (Q
a
),
total scattering (Q
s
), and forward scattering
(Q
f
) efciencies of liquid water drops of dif-
ferent sizes at = 0.50 m (n
= 1.335,
=
1.0 10
9
). The efciency for water at any
other wavelength
1
in the visible or UV spec-
trum and diameter d
1
is the efciency in the
gure located at diameter d = 0.5md
1
1
.
When a particles diameter is near the wavelength of light (0.03 - d
i
- 32 or
0.1 -
i.
- 100), the particle is in the Mie regime. In this regime, absorption and
scattering efciencies are approximated with Mies solution to Maxwells equa-
tions. The scattering efciency of a particle in the Mie regime is determined by
Q
s.i.
=
2
i.
k=1
(2k +1)(|a
k
|
2
+|b
k
|
2
) (9.47)
307
where a
k
and b
k
are complex functions (e.g., van de Hulst 1957; Kerker 1969;
Toon and Ackerman 1981; Bohren and Huffman 1983; Liou 2002). The single-
particle absorption efciency of a particle in the Mie regime is estimated from
Q
a.i.
= Q
e.i.
Q
s.i.
, where
Q
e.i.
=
2
i.
k=1
(2k +1)Re(a
k
+b
k
) (9.48)
is the single-particle total extinction efciency in the Mie regime. Figures 9.20 and
9.21 show Q
a.i.
and Q
s.i.
for soot and liquid water, as a function of particle size
at = 0.50 m. The gures also show the forward scattering efciency Q
f.i.
,
which is the efciency with which a particle scatters light in the forward direction.
The forward scattering efciency is always less than the total scattering efciency.
The proximity of Q
f.i
to Q
s.i
in both gures indicates that aerosol particles scat-
ter strongly in the forward direction. The difference between Q
s.i
and Q
f.i
is the
scattering efciency in the backward direction.
Figure 9.20 shows that visible-light absorption efciencies of soot particles peak
when the particles are 0.2 to 0.4 m in diameter. Such particles are in the accumu-
lation mode (0.12 m in diameter, Chapter 13) with respect to particle size and in
the Mie regime with respect to the ratio of particle size to the wavelength of light.
Figure 9.21 shows that water particles 0.3 to 2 m in diameter scatter visible light
more efciently than do smaller or larger particles. These particles are also in the
accumulation mode with respect to particle size and in the Mie regime with respect
to the ratio of particle size to the wavelength of light.
Because the accumulation mode contains a relatively high particle number con-
centration, and because particles in this mode have high scattering and absorption
(with respect to soot) efciencies, the accumulation mode almost always causes
more light reduction than do smaller or larger modes (Waggoner et al. 1981). In
many urban regions, 20 to 50 percent of the accumulation mode mass is sulfate.
Thus, sulfate is correlated with particle scattering more closely than is any other
particulate species, aside from liquid water.
When a particles diameter is much larger than the wavelength of light (d
i
>
32 or
i.
> 100), the particle is in the geometric regime. Such particles reect,
refract, and diffract light signicantly. As
i.
increases from the Mie regime to the
geometric regime, the scattering efciency approaches a constant in a sinusoidal
fashion, as shown in Fig. 9.21. As
i.
, the scattering efciency approaches
another constant, given by Chylek (1977) as
lim
i.
Q
s.i.
= 1 +
1
m
+2
(9.49)
Although cloud drops, which have diameters >5 m, have lower scattering ef-
ciencies and a lower number concentration than do accumulation-mode aerosol
particles, clouds can obscure visibility to a greater extent than can aerosol parti-
cles because cloud drops have much larger cross-sectional areas than do aerosol
particles.
308
Example 9.8
At =0.5 m, calculate the scattering efciency of liquid water as
i.
.
SOLUTION
From Table 9.6, n
= 1.335 at = 0.5 m for liquid water. From (9.49), Q

s.i.
=
1.1. Hence, the scattering efciency at =0.5 m, showninFig. 9.21, decreases
from 2 to 1.1 as the diameter approaches innity. The absorption efciency
also approaches 1.1 at innite diameter.
Deirmendjian (1969) showed that, as
i.
, a particles absorption efciency
converges to its scattering efciency regardless of howweak the imaginary index of
refraction is. Since all liquids and solids scatter in the geometric regime, they must all
absorb at large enough size, as well. In other words, large particles absorb radiation,
regardless of their imaginary index of refraction. Figure 9.21, for example, shows
how the absorption efciency of liquid water increases with increasing diameter
even when
= 1.0 10
9
. The gure also shows that the absorption efciency
becomes important only when liquid drops reach raindrop-size (>1000 m in
diameter). The absorptivity of raindrops causes the bottoms of precipitating clouds
to appear gray or black.
9.4.3.4 Optics of particle mixtures
In the previous discussion, particles were assumed to be externally mixed, or in
isolation from one another. In reality, particles are often internally mixed, meaning
they contain multiple components. Several methods have been developed to treat
the effective refractive index of a mixture. These treatments are called effective
medium approximations. A more detailed discussion and comparison of approxi-
mations is given in Bohren and Huffman (1983) and Chylek et al. (1988).
The simplest way to account for the internal mixing of particle components
in a model is to average the complex refractive index of each component within
a particle by volume. This method is called the volume average refractive index
mixing rule and is quantied with
m
=
N
V
q=1
m
.q
=
N
V
q=1
(n
) (9.50)
where N
V
is the number of volume components in a particle, is the volume of
a single particle (cm
3
particle
1
), and
q
is the volume of component q within the
particle (cm
3
particle
1
). The ratio
q
is the volume fraction of component q
within the particle.
A second method is to average the dielectric constant of each component by
volume (e.g., Chylek et al. 1988). A dielectric is a nonconducting material that
can sustain a steady electric eld and serve as an insulator. The dielectric constant
309
(also called the permittivity) is a property of a material that determines how much
electrostatic energy can be stored per unit volume of the material when unit voltage
is applied. It is the ratio of the capacitance of a capacitor with the material to the
same capacitor when the dielectric is a vacuum. The complex dielectric constant is
related to the complex refractive index of a material by
= m
2
= (n
)
2
= n
2
i2n
=
r.
i
i.
(9.51)
where
r.
= n
2

i.
= 2n
(9.52)
are the real and imaginary components, respectively, of the complex dielectric
constant. The real and imaginary refractive indices are related to these components
by
n
2
r.
+
2
i.
+
r.
2

2
r.
+
2
i.
r.
2
(9.53)
respectively. The volume average dielectric constant mixing rule is
m
2
=
N
V
q=1
m
2
.q
=
N
V
q=1
.q
=
N
V
q=1
(
r.
i
i.
) (9.54)
A third method is the Maxwell Garnett mixing rule (Maxwell Garnett 1904).
This rule gives the effective complex refractive index of absorbing inclusions (sub-
script A) embedded within a matrix of a single homogeneous substance (subscript
M) as
m
2
=
.M
1 +
3
N
A
q=1
A.q

.A.q
.M
.A.q
+2
.M
1
N
A
q=1
A.q

.A.q
.M
.A.q
+2
.M
(9.55)
where N
A
is the number of absorbing substances embedded in the matrix,
.A.q
is
the complex dielectric constant of each absorbing substance, and
.M
is the com-
plex dielectric constant of the mixture. Equation (9.55) is solved by rst substituting
= n
2
2
i2n from (9.51) into each complex dielectric constant on the right
side of the equation, rewriting the equation in the form
= (a ib)(c id),
then multiplying through by (c +id)(c +id) to obtain
=
r.
i
i.
, where
r.
= (ac +bd)(c
2
d
2
) and
i.
= (bc ad)(c
2
d
2
). Finally, the average real
and imaginary refractive indices are obtained from
r.
and
i.
from (9.53).
A related mixing rule is the inverted Maxwell Garnett mixing rule, which is the
same as (9.55) but with the subscripts M and A reversed in all terms.
310
A fourth method is the Bruggeman mixing rule (Bruggeman 1935), which is
derived from the same base equation as the Maxwell Garnett mixing rule, but
using a different approximation, and also assumes one or more absorbing inclusions
embedded within a matrix. With this rule, the effective complex dielectric constant,
, is solved iteratively from

N
A
q=1
A.q

.A.q
.A.q
+2
1
N
A
q=1
A.q

.M
.M
+2
= 0 (9.56)
The complex refractive index is then obtained from
with (9.51).
All the effective medium theories discussed result in an average complex refrac-
tive index for an entire particle. Optical properties of the particle are then calculated
fromthis average refractive index. However, when an absorbing solid, such as soot,
exists in an otherwise liquid particle, the soot is a distinct component and not dis-
tributed equally throughout the particle. The optical properties of the particle in
this case depend on radiative interactions between the liquid and solid, which the
effective medium approximations do not capture.
A more physical, but still somewhat limited, way of treating absorption in a
multicomponent particle is to treat the absorbing component as a core material
surrounded by a shell with a volume-averaged refractive index. Theoretical cal-
culations with this method suggest that, when an absorbing particle component,
such as soot, becomes coated by a relatively nonabsorbing material, such as sul-
furic acid or organic carbon, the absorption efciency of the soot increases (e.g.,
Toon and Ackerman 1981; Jacobson 1997a,b, 2000, 2001b; Lesins et al. 2002).
Figure 9.22 shows the theoretical enhancement of absorption due to the addition of
a coating onto particles containing black carbon. The theory has been supported
experimentally by Schnaiter et al. (2003), who found that the addition of just
a few monolayers of a condensed organic coating to a soot particle increased its
absorption coefcient by 35 percent. Subsequent experiments showed that conden-
sation of more layers increased absorption by 200 percent, a nding consistent with
Fig. 9.22. The enhancement of absorption due to the coating of soot in the atmo-
sphere may be sufciently great to make soot the second most important component
of global warming, after CO
2
and ahead of CH
4
, in terms of its direct radiative
effect (Jacobson 2000, 2001b).
The reasons for the enhanced absorption of a soot particle when it is coated
versus when it is not are as follows: (1) In the case of particles larger than the
wavelength of light, geometric optics implies that n
2
more light is incident on a

small sphere when it is at the center of a much larger transparent sphere with real
refractive index n
than when it is in air (e.g., Bohren 1986; Twohy et al. 1989).

This is called the optical focusing effect, because more light is focused into the core
of a sphere due to refraction when a shell surrounds the core compared with when
it does not. (2) In the case of particles near to or smaller than the wavelength of
light, enhanced diffraction at the edge of a particle increases the exposure of the
core to waves in comparison with exposure of the core to waves in the absence of
a shell.
311
0
1
2
3
4
5
6
0.01 0.1 1 10 100
0. 04
0. 08
0. 12
Total particle (soot core+S(VI) shell) diameter (m)
A
b
s
o
r
p
t
i
o
n

c
r
o
s
s

s
e
c
t
i
o
n
e
n
h
a
n
c
e
m
e
n
t

f
a
c
t
o
r
d
BC
(m)
Figure 9.22 Absorption cross section enhancement
factors at = 0.5m, due to coating a black car-
bon (BC) core with a shell comprised of sulfu-
ric acidwater (S(VI)). The enhancement factor is
the ratio of the absorption cross section of a total
(coated) particle with a BC core (either 40, 80, or
120 nm in diameter) and an S(VI) shell to the larger
of the absorption cross section of BCalone (at either
40, 80, or 120 nm in diameter) or of pure S(VI)
at the same size as the total core + shell particle
(which affects the factor only at large sizes). The
leftmost values in the gure are the enhancement
factors (1.0) corresponding to the pure core. The
rightmost values (1.0) are those corresponding to a
large pure S(VI) drop that absorbs since all particles
with a nonzero imaginary refractive index become
absorbing at some size (Section 9.4.3.3).
Although treating soot as a concentric core is not always realistic because soot is
an amorphous, randomly shaped agglomerate of many individual spherules, such
treatment becomes more realistic as more material condenses. The reason is that
soot aggregates collapse under the weight of a heavy coating, becoming more spher-
ical in the process (Schnaiter et al. 2003; Wentzel et al. 2003). Other techniques
of treating soot inclusions within particles, aside from the concentric core treat-
ment, include treating a composite of several whole soot particles distributed in
an internal mixture (applicable to cloud drops) (Chylek et al. 1988), embedding
soot inclusions at random locations in a particle (Chylek et al. 1995; Fuller 1995),
treating soot as a subsurface inclusion or grain on a particle (Fuller et al. 1999),
and treating soot as a coated chain aggregate (Fuller et al. 1999).
9.4.3.5 Modeled aerosol extinction
Figures 9.23 (a) and (b) show modeled vertical proles of extinction coefcients
over a polluted airshed. Both gures indicate that extinction coefcients increased
fromthe free troposphere (altitudes above 900 hPa) to the boundary layer. The rst
gure indicates that, at = 0.32 m (ultraviolet light), extinction due to aerosol
scattering in the boundary layer was less than that due to aerosol absorption. At
312
9.5 Visibility
0 0.1 0.2 0.3 0.4 0.5
300
400
500
600
700
800
900
1000
0 0.1 0.2 0.3 0.4 0.5
(km
1
)
P
r
e
s
s
u
r
e

(
h
P
a
)
Gas scat.
Aer. scat.
Aer. abs. Gas abs.
(a)
0 0.1 0.2 0.3 0.4 0.5
300
400
500
600
700
800
900
1000
0 0.1 0.2 0.3 0. 4 0.5
(km
1
)
P
r
e
s
s
u
r
e

(
h
P
a
)
Gas scat.
Aer. scat.
Aer. abs.
Gas abs.
(b)
Figure 9.23 Modeled proles from Claremont, California (11.30 on August
27, 1997) of extinction coefcients (km
1
) due to gas scattering, aerosol
scattering, gas absorption, and aerosol absorption at (a) = 0.32 and
(b) = 0.61m. The peaks at 920 hPa indicate elevated aerosol layers. From
Jacobson (1998b).
= 0.61 m (visible light), the reverse was true because the aerosol absorptivity
decreased from the ultraviolet to visible spectra. Above the boundary layer, gas
scattering dominated aerosol scattering at 0.32 and 0.61 m.
9.5 VISIBILITY
A result of aerosol buildup is visibility degradation. Although unnatural visibility
degradation, itself, causes no adverse health effects, it usually indicates the presence
of pollutants, which are often harmful.
Visibility is a measure of how far we can see through the air. Even in the cleanest
air, our ability to see along the Earths horizon is limited to a fewhundred kilometers
by background gases and aerosol particles. If we look up through the sky at night,
however, we can discern light from stars that are millions of kilometers away. The
difference between looking horizontally and vertically is that more gas molecules
and aerosol particles lie in front of us in the horizontal than in the vertical.
Several terms describe maximumvisibility. Two subjective terms are visual range
and prevailing visibility. Visual range is the actual distance at which a person can
discern an ideal dark object against the horizon sky. Prevailing visibility is the great-
est visual range a person can see along 50 percent or more of the horizon circle
(360
), but not necessarily in continuous sectors around the circle. It is determined

by a person who identies landmarks known distances away in a full 360
circle
around an observation point. The greatest visual range observed over 180
or more
of the circle (not necessarily in continuous sectors) is the prevailing visibility. Thus,
half the area around an observation point may have visibility worse than the pre-
vailing visibility, which is important because most prevailing visibility observations
are made at airports. If the visual range in a sector is signicantly different from
313
x
0
x
dx
Scattering out of path
Scattering into path
I
0
I
Figure 9.24 Change of radiation intensity along a
beam. A radiation beam originating from a dark object
has intensity I = 0 at point x
0
. Over a distance dx, the
beams intensity increases due to scattering of back-
ground light into the beam. This added intensity is
diminished somewhat by absorption along the beam
and scattering out of the beam. At point x, the net
intensity of the beam has increased close to that of the
background intensity.
the prevailing visibility, the observer at an airport usually denotes this information
in the observation record.
A less subjective and, now, a regulatory denition of visibility is the meteo-
rological range. Meteorological range can be explained in terms of the following
example. Suppose a perfectly absorbing dark object lies against a white background
at a point x
0
, as shown in Fig. 9.24. Because the object is perfectly absorbing, it
reects and emits no visible radiation; thus, its visible radiation intensity (I) at
point x
0
is zero, and it appears black. As a viewer backs away from the object,
background white light of intensity I
B
scatters into the eld of view, increasing the
intensity of light in the viewers line of sight. Although some of the added back-
ground light is scattered out of or absorbed along the eld of view by gases and
aerosol particles, at some distance away fromthe object, so much background light
has entered the path between the viewer and the object that the viewer can barely
discern the black object against the background light.
The meteorological range is a function of the contrast ratio, dened as
C
ratio
=
I
B
I
I
B
(9.57)
The contrast ratio gives the difference between the background intensity and the
intensity in the viewers line of sight, all relative to the background intensity. If the
contrast ratio is unity, then an object is perfectly visible. If it is zero, then the object
cannot be differentiated from background light.
The meteorological range is the distance from an object at which the contrast
ratio equals the liminal contrast ratio of 0.02 (2 percent). The liminal or threshold
contrast ratio is the lowest visually perceptible brightness contrast a person can see.
It varies fromindividual to individual. Koschmieder (1924) selected a value of 0.02.
Middleton (1952) tested 1000 people and found a threshold contrast range of
314
9.5 Visibility
between 0.01 and 0.20, with the mode of the sample between 0.02 and 0.03.
Campbell and Maffel (1974) found a liminal contrast of 0.003 in laboratory studies
of monocular vision. Nevertheless, 0.02 has become an accepted liminal contrast
value for meteorological range calculations. In sum, the meteorological range is the
distance from an ideal dark object at which the object has a 0.02 liminal contrast
ratio against a white background.
The meteorological range can be derived from the equation for the change in
object intensity along the path described in Fig. 9.24. This equation is
dI
dx
=
t
(I
B
I) (9.58)
where all wavelength subscripts have been removed,
t
is the total extinction coef-
cient,
t
I
B
accounts for the scattering of background light radiation into the path,
and
t
I accounts for the attenuation of radiation along the path due to scattering
out of the path and absorption along the path. A total extinction coefcient is the
sumof extinction coefcients due to scattering and absorption by gases and aerosol
particles. Thus,
t
=
a.g
+
s.g
+
a.a
+
s.a
(9.59)
Integrating Equation (7.9) from I = 0 at point x
0
= 0 to I at point x with constant
t
yields the equation for the contrast ratio,
C
ratio
=
I
B
I
I
B
= e
t
x
(9.60)
When C
ratio
= 0.02 at a wavelength of 0.55 m, the resulting distance x is the
meteorological range (also called the Koschmieder equation).
x =
3.912
t
(9.61)
In polluted tropospheric air, the only important gas-phase visible-light attenu-
ation processes are Rayleigh scattering and absorption by nitrogen dioxide (Wag-
goner et al. 1981). Several studies have found that scattering by particles, par-
ticularly those containing sulfate, organic carbon, and nitrate, may cause 6095
percent of visibility reduction and absorption by soot may cause 540 percent of
visibility reduction in polluted air (Cass 1979; Tang et al. 1981; Waggoner et al.
1981).
Table 9.8 shows meteorological ranges derived from extinction coefcient mea-
surements for a polluted and less-polluted day in Los Angeles. Particle scattering
dominated light extinction on both days. On the less-polluted day, gas absorption,
particle absorption, and gas scattering all had similar small effects. On the pol-
luted day, the most important visibility reducing processes were particle scattering,
particle absorption, gas absorption, and gas scattering, in that order.
315
Table 9.8 Meteorological ranges (km) resulting from gas scattering, gas absorption,
particle scattering, particle absorption, and all processes at a wavelength of 0.55m
on a polluted and less-polluted day in Los Angeles
Gas Gas Particle Particle
Day scattering absorption scattering absorption All
Polluted (8/25/83) 366 130 9.6 49.7 7.42
Less-polluted (4/7/83) 352 326 151 421 67.1
Meteorological ranges derived from extinction coefcients of Larson et al. (1984).
9.6 OPTICAL DEPTH
Apurpose of simulating atmospheric radiation is to estimate the spectral irradiance
(F
) that reaches a given layer of the atmosphere. The rst step in such a calcu-
lation is to determine the total spectral extinction coefcient at each wavelength
with
=
s.g.
+
a.g.
+
s.a.
+
a.a.
+
s.h.
+
a.h.
(9.62)
where the individual coefcients are for scattering by gases, absorption by gases,
scattering by aerosol particles, absorption by aerosol particles, scattering by
hydrometeor particles, and absorption by hydrometeor particles, respectively. The
next step is to dene the vertical component of the incremental distance along a
beam of interest through which radiation travels. When the beam of interest is the
solar beam, this component is
dz = cos
s
dS
b
= j
s
dS
b
(9.63)
where dS
b
is the incremental distance along the beam,
s
is the solar zenith angle,
and
j
s
= cos
s
(9.64)
The solar zenith angle is the angle between the surface normal (a line directed from
the center of the Earth that extends vertically above the surface) and the direction
of the Sun. Figure 9.25 shows the relationship among dz, dS
b
, and
s
.
Integrating the spectral extinction coefcient over an incremental distance gives
an optical depth (dimensionless). For radiative transfer calculations, the distance
of interest is in the vertical. In such cases, optical depth quanties scattering and
absorbing that occurs between the top of the atmosphere and a given altitude. The
optical depth increases from zero at the top of the atmosphere to a maximum at
the ground. The incremental optical depth is
d
dz =
j
s
dS
b
(9.65)
which increases in the opposite direction from incremental altitude. Integrating
(9.65) from the top of the atmosphere (z = S
b
= ) to any altitude z, which
316
9.7 Solar zenith angle
Sun
dz
dt
dS
b
q
s
t = t z = 0 S
b
= 0
t = 0

z = S
b

=
Figure 9.25 Relationship among incre-
mental optical depth (d), incremental
altitude (dz), solar zenith angle (
s
), and
incremental distance along the beam
(dS
b
).
corresponds to a location S
b
along the beam of interest, gives the optical depth as
dz =
S
b
j
s
dS
b
(9.66)
9.7 SOLAR ZENITH ANGLE
The solar zenith angle can be determined from
cos
s
= sin sin +cos cos cos H
a
(9.67)
where is the latitude, is the solar declination angle, and H
a
is the local hour
angle of the Sun, as illustrated in Figs. 9.26 (a) and (b). The declination angle is the
angle between the Equator and the north or south latitude of the subsolar point,
which is the point at which the Sun is directly overhead. The local hour angle is
the angle, measured westward, between the longitude (meridian) of the subsolar
point and the longitude of the location of interest.
Equation (9.67) is obtained from Fig. 9.26(a) by applying the law of cosines
to triangle APB. Since arcangle AOP = 90
, arcangle BOP = 90
, and the
distances OP, OA, and OB are known, the arc lengths AP and BP can be determined
from the law of sines. With these distances and the fact that angle APB = H
a
, the
arc length AB can be obtained from the law of cosines. Since the distances AB, OA,
and OB are known, the solar zenith angle (AOB) can be determined from the law
of sines (e.g., Hartman 1994).
The solar declination angle is found from
= sin
1
(sin
ob
sin
ec
) (9.68)
where
ob
is the obliquity of the ecliptic and
ec
is the ecliptic longitude of the Sun.
The ecliptic is the mean plane of the Earths orbit around the Sun. It is xed in
317
Surface
normal
Sun
A
B
C
D
O
P
N
Surface
normal
Sun
d
H
a
j
q
s
(a) (b)
Figure 9.26 (a) Geometry for zenith angle calculations on a sphere. The ray OAN is
the surface normal above the point of interest. The ray OB points to the Sun, which
is incident directly over point B. Point B is the subsolar point. Angle AOB is the solar
zenith angle (
s
). Angle BOD is the solar declination (), angle AOC is the latitude
() of the surface normal, angle CAB is the solar azimuth angle (
s
), and angles COD
= CPD = APB are hour angles (H
a
). (b) Geometry for a different solar zenith angle.
space relative to the Earth, and the Earth rotates through it. It cuts through the
Tropic of Capricorn on one side of the Earth, the Equator in the middle, and the
Tropic of Cancer on the other side of the Earth. The obliquity of the ecliptic is
the angle between the plane of the Earths Equator and the plane of the ecliptic,
approximated as
ob
= 23
.439 0
.0000004N
JD
(9.69)
(NAO 1993), where
N
JD
= 364.5 +(Y 2001) 365 + D
L
+ D
J
(9.70)
D
L
=
INT(Y 2001)4 Y 2001

INT(Y 2000)4 1 Y - 2001
is the number of days from the beginning of Julian year 2000. In (9.70), Y is the
current year, D
L
is the number of leap days since or before the year 2000, D
J
is
the Julian day of the year, which varies from 1 on January 1 to 365 (for nonleap
years) or 366 (for leap years) on December 31. Leap years occur every year evenly
divisible by 4.
The ecliptic longitude of the Sun is approximately
ec
= L
M
+1
.915 sin g
M
+0
.020 sin 2g
M
(9.71)
where
L
M
= 280
.460 +0
.9856474N
JD
g
M
= 357
.528 +0
.9856003N
JD
(9.72)
are the mean longitude of the Sun and the mean anomaly of the Sun, respectively.
The mean anomaly of the Sun is the angular distance, as seen by the Sun, of the
318
9.7 Solar zenith angle
N
.
H
.

w
i
n
t
e
r
,

S
.
H
.

s
u
m
m
e
r
s
o
l
s
t
i
c
e
,

D
e
c
.

2
2
23.5
23.5
Vernal equinox, Mar. 20

Autumnal equinox, Sept. 23
Tropic of
Cancer
Topic of
Capricorn
Earth
Sun
Sun
Sun
Equator
Axis of
rotation
N
.
H
.

s
u
m
m
e
r
,

S
.
H
.

w
i
n
t
e
r
s
o
l
s
t
i
c
e
,

J
u
n
e

2
2
Figure 9.27 Solar declination angles during solstices
and equinoxes. Of the four times shown, the EarthSun
distance is greatest at the summer solstice.
Earth from its perihelion, which is the point in the Earths orbit at which the Earth
is closest to the Sun assuming the Earths orbit is perfectly circular and the Earth
is moving at a constant speed. The mean anomaly at the perihelion is 0
.
The local hour angle (in radians) is
H
a
=
2t
s
86 400
(9.73)
where t
s
is the number of seconds past local noon, and 86 400 is the number
of seconds in a day. At noon, when the Sun is highest, the local hour angle is
zero, and (9.67) simplies to cos
s
= sin sin +cos cos . When the Sun is over
the Equator, the declination angle and latitude are zero, and (9.67) simplies to
cos
s
= cos H
a
. Figure 9.27 shows that the Sun reaches its maximum declination
(23.5
) at the summer and winter solstices and its minimum declination (0
) at
the vernal and autumnal equinoxes.
Example 9.9
Calculate the solar zenith angle at 1:00 p.m. PST, on February 27, 1994 at a
latitude of = 35
N.
SOLUTION
February 27 corresponds to Julian day D
J
= 58. From (9.70), N
JD
= 2134.5.
From (9.72), L
M
= 1823.40
and g
M
= 1746.23
. Thus, from (9.71), (9.69),

and(9.68),
ec
=1821.87
,
ob
= 23.4399
, and =8.52
, respectively. Equa-
tion (9.73) gives H
a
=15.0
. Thus, from (9.67), cos

s
= sin35
sin(8.52
) +
cos 35
cos(8.52
) cos 15.0

s
= 45.8
.
319
Sun
(m, f)
(m', f')
(m
s
, f
s
)
F
s
Multiple
scattering
Diffuse
Single
scattering
Direct
Beam
Figure 9.28 Single scattering of direct solar radi-
ation and multiple scattering of diffuse radiation
adds to the intensity along a beam of orientation j,
. The parameter j = cos is always positive, but
when a ray is directed upward, +jis used, and when
a ray is directed downward, j is used. Adapted
from Liou (2002).
9.8 THE RADIATIVE TRANSFER EQUATION
The radiative transfer equation gives the change in radiance and/or irradiance along
a beam of electromagnetic energy at a point in the atmosphere. Radiances are used
to calculate actinic uxes, which are used in photolysis coefcient equations. Irradi-
ances are used to calculate heating rates, which are used in temperature calculations.
The processes affecting radiation along a beam are scattering of radiation out of
the beam, absorption of radiation along the beam, multiple scattering of indirect,
diffuse radiation into the beam, single scattering of direct, solar radiation into the
beam, and emission of infrared radiation into the beam. Single scattering occurs
when a photon of radiation is redirected into a beam after it collides with a particle
or gas molecule, as shown in Fig. 9.28. Multiple scattering occurs when a photon
enters a beam after colliding sequentially with several particles or gas molecules,
each of which redirects the photon. Solar radiation that has not yet been scattered
is direct radiation. Radiation, either solar or infrared, that has been scattered is
diffuse radiation.
The change in spectral radiance over the distance dS
b
along a beam is
dI
= dI
so.
dI
ao.
+dI
si.
+dI
Si.
+dI
ei.
(9.74)
where
dI
so.
= I
s.
dS
b
(9.75)
320
9.8 The radiative transfer equation
represents scattering of radiation out of the beam,
dI
ao.
= I
a.
dS
b
(9.76)
represents absorption of radiation along the beam,
dI
si.
=
s.k.
4
2
0
1
1
I
.j
.
P
s.k..j.j
..
dj
dS
b
(9.77)
represents multiple scattering of diffuse radiation into the beam,
dI
Si.
=
s.k.
4
P
s.k.j.j
s
..
s
F
s.
e
j
s
dS
b
(9.78)
represents single scattering of direct solar radiation into the beam, and
dI
ei.
=
a.
B
.T
dS
b
(9.79)
represents emission of infrared radiation into the beam. In (9.77) and (9.78), the
summations are over all scattering processes (k =g for gases, a for aerosol particles,
and h for hydrometeor particles), and the P
s
factors are scattering phase functions,
to be dened shortly. In (9.75), (9.76), and (9.79), the factors,
s.
=
s.g.
+
s.a.
+
s.h.

a.
=
a.g.
+
a.a.
+
a.h.
(9.80)
are extinction coefcients due to total scattering and absorption, respectively. The
extinction coefcient due to total scattering plus absorption is
=
s.
+
a.
(9.81)
In the solar spectrum, single scattering of solar radiation is more important than
is radiative emission. In the infrared spectrum, the reverse is true. Thus, (9.78) is
used for solar wavelengths and (9.79) is used for infrared wavelengths.
9.8.1 Phase function and asymmetry parameter
In (9.77), P
s.k..j.j
..
is the scattering phase function, which gives the angular
distribution of scattered energy as a function of direction. It relates how diffuse
radiation, which has direction j
, is redirected by gases or particles toward the

beamof interest, which has direction j, , as shown in Fig. 9.28. In the gure, j
=
cos
and j = cos , where
and are the zenith angles of the diffuse radiation

and the beam of interest, respectively. Similarly,
and are the azimuth angles of

the diffuse radiation and the beam of interest, respectively. The integral in (9.77) is
over all possible angles of incoming multiple-scattered radiation. Scattering phase
functions vary with wavelength and differ for gases, aerosol particles, and cloud
drops.
In (9.78), P
s.k..j.j
s
..
s
is the scattering phase function for direct radiation. The
function relates how direct solar radiation, with direction j
s
(= cos
s
),
s
, is
redirected by gases or particles to j, , as shown in Fig. 9.28. This phase function
is not integrated over all solid angles, since single-scattered radiation originates
from one angle.
321
1.5
1
0.5
0
0.5
1
1.5
2 1.5 1 0.5 0 0.5 1 1.5 2
() P
s
1.5
1
0.5
0
0.5
1
1.5
2 1.5 1 0.5 0 0.5 1 1.5 2
() P
s
(a) (b)
Figure 9.29 Scattering phase function distribution in polar coordinates for
(a) isotropic and (b) Rayleigh scattering. The diagrams were generated from (9.84)
and (9.85), respectively.
The scattering phase function is dened such that
1
4
4
P
s.k.
(O) dO
a
= 1 (9.82)
In this equation, O is the angle between directions j
and j, , as shown in
Fig. 9.28. Thus, P
s.k..j.j
..
= P
s.k.
(O). Here, O
s
is the angle between the solar
beam (j
s
.
s
) and the beam of interest (j. ). Thus, P
s.k..j.j
s
..
s
= P
s.k.
(O
s
).
Substituting dO
a
= sinOdOd from (9.8) into (9.82) gives
1
4
2
0

0
P
s.k.
(O) sinOdOd = 1 (9.83)
The integral limits are dened so that the integration is over a full sphere. For
isotropic scattering, the phase function is
P
s.k.
(O) = 1 (9.84)
which satises (9.82). For Rayleigh (gas) scattering, the phase function is
P
s.k.
(O) =
3
4
(1 +cos
2
O) (9.85)
which also satises (9.82). Figures 9.29 (a) and (b) showscattering phase functions
for isotropic and Rayleigh scattering, respectively, versus scattering angle. The
phase function for isotropic scattering projects equally in all directions, and that for
Rayleigh scattering is symmetric, but projects mostly in the forward and backward
directions.
An approximation to the phase function for Mie scattering by aerosol parti-
cles and hydrometeor particles is the HenyeyGreenstein function (Henyey and
Greenstein 1941),
P
s.k.
(O) =
1 g
2
a.k.
1 + g
2
a.k.
2g
a.k.
cos O
32
(9.86)
322
where g
a.k.
is the asymmetry parameter, dened shortly. This equation is valid
primarily for scattering that is not strongly peaked in the forward direction (Liou
2002), and requires advance knowledge of the asymmetry parameter, which is a
function of the phase function itself.
The asymmetry parameter, or rst moment of the phase function, is a parameter
derived from the phase function that gives the relative direction of scattering by
particles or gases. Its analytical form is
g
a.k.
=
1
4
4
P
s.k.
(O) cos OdO
a
(9.87)
The asymmetry parameter approaches +1 for scattering strongly peaked in the
forward direction and 1 for scattering strongly peaked in the backward direction.
If the asymmetry parameter is zero, scattering is equal in the forward and backward
directions. In sum,
g
a.k.
> 0 forward (Mie) scattering

= 0 isotropic or Rayleigh scattering
- 0 backward scattering
(9.88)
Expanding (9.83) with dO
a
= sinOdOd yields
g
a.k.
=
1
4
2
0

0
P
s.k.
(O) cos OsinOdOd (9.89)
For isotropic scattering, where P
s.k.
(O) = 1. the asymmetry parameter simpli-
es to
g
a.k.
=
1
4
2
0

0
cos OsinOdOd =
1
2
1
1
jdj = 0 (9.90)
where j = cos O. The zero asymmetry parameter for isotropic scattering is
expected, since isotropic scattering distributes radiation equally in all directions.
Substituting the phase function for Rayleigh scattering into (9.85) gives
g
a.k.
=
1
4
2
0

0
3
4
(1 +cos
2
O) cos OsinOdOd
=
3
8
2
0
1
1
(j +j
3
)djd = 0 (9.91)
The Rayleigh scattering intensity is evenly distributed between forward and back-
ward directions, but Rayleigh scattering is not isotropic, since the radiation is not
scattered equally in all directions.
The asymmetry parameter for aerosol particles and hydrometeor particles can
be found with
g
a.k.
= Q
f.i.
Q
s.i.
(9.92)
323
where Q
f.i.
is the single-particle forward scattering efciency and Q
s.i.
is the
single-particle total scattering efciency, both of which can be obtained from a
Mie-scattering algorithm. Curves for both parameters were shown in Figs. 9.20
and 9.21 for two substances. These gures indicate that the asymmetry parameters
for soot and liquid water aerosol particles are typically between 0.6 and 0.85.
Measured asymmetry parameters in clouds range from 0.7 to 0.85 (e.g., Gerber
et al. 2000).
Another parameter derived from the phase function is the backscatter ratio,
b
a.k.
=
2
0
2
P
s.k.
(O) sinOdOd
2
0
0
P
s.k.
(O) sinOdOd
(9.93)
This parameter gives the fraction of total scattered radiation that is scattered in
the backward hemisphere of the particle. Typical values of the backscatter ratio
are 0.080.2 (e.g., Marshall et al. 1995), suggesting that 8092 percent of energy
scattered by particles is scattered in the forward direction. Particle shape has a
signicant impact on the backscatter ratio through its effect on the phase function
(e.g., Pollack and Cuzzi 1980; Fridlind and Jacobson 2003).
9.8.2 Incident solar radiation
In (9.78), F
s.
is the incident solar radiation at the top of the atmosphere in a wave-
length interval centered at . The radiation intensity at the top of the atmosphere
depends on the solar luminosity and the EarthSun distance. The solar luminosity
(L
p
) is the total energy per unit time emitted by the Suns photosphere and is
approximately 3.9 10
26
W. The irradiance, or luminosity emitted per unit area
(W m
2
) at the photosphere, is approximately
F
p
=
L
p
4 R
2
p
=
B
T
4
p
(9.94)
where R
p
=6.96 10
8
mis the radius of the Sun, fromits center to the photosphere,
T
p
is the temperature of the Sun at the photosphere, and (9.94) is a summation
over all wavelength intervals.
Example 9.10
For a photosphere temperature of T
p
= 5796 K, F
p
6.4 10
7
W m
2
from
(9.94).
Irradiance decreases proportionally to the inverse square distance from its
source. The yearly mean solar irradiance reaching the top of Earths atmosphere,
324
Obliquity
23.5
Sun
Autumnal equinox
Sept. 23
Vernal equinox
Mar. 20
147 million km 152 million km
N.H. winter
S.H. summer
solstice
Dec. 22
N.H. summer
S.H. winter
solstice
June 22
Figure 9.30 Relationship between the Sun and Earth during the
solstices and equinoxes.
or solar constant, is related to F
p
by
F
s
=
R
p
R
es
2
F
p
=
R
p
R
es
B
T
4
p
(9.95)
where

R
es
= 1.5 10
11
m is the mean distance from the center of the Sun to
the Earth. Equation (9.95), with the parameters specied above, gives

F
s
1379
W m
2
, which is close to measurements of

F
s
= 1365 W m
2
. The value of

F
s
varies by 1 W m
2
over each 11-year sunspot cycle.
The actual total irradiance at the top of the atmosphere at a given time is
F
s
=
R
es
R
es
F
s
(9.96)
where R
es
is the actual distance between the Sun and the Earth on a given day, and
R
es
R
es
2
1.00011 +0.034221 cos
J
+0.00128 sin
J
+0.000719 cos 2
J
+0.000077 sin 2
J
(9.97)
(Spencer 1971). In (9.97),
J
= 2D
J
D
Y
, where D
Y
is the number of days in a
year (365 for nonleap years and 366 for leap years), and D
J
is the Julian day of
the year. Figure 9.30 shows that the Earth is further from the Sun in the Northern
Hemisphere summer than in the winter by about 3.4 percent.
Example 9.11
On December 22, the total irradiance at the top of Earths atmosphere is F
s
=
1365 1.034 = 1411 W m
2
, and on June 22, the irradiance is F
s
= 1365
0.967 = 1321 W m
2
. Thus, the irradiance varies by 90 W m
2
(6.6 percent)
between December and June.
In a model, the actual total solar irradiance reaching the top of Earths atmo-
sphere (F
s
) (W m
2
) is related to the actual solar irradiance in each wavelength
325
interval (F
s.
) (Wm
2
m
1
), the mean solar irradiance in each wavelength interval
(

F
s.
) (W m
2
m
1
), and the solar constant (

F
s
) (W m
2
) by
F
s
=
(F
s.
L) =
R
es
R
es

F
s.
L
R
es
R
es
F
s
(9.98)
where L is a wavelength interval (m). Values for

F
s.
are given in Appendix
Table B.2.
The solar constant

F
s
can be used to determine Earths equilibrium temperature
in the absence of the greenhouse effect. The total energy (W) absorbed by the
Earthatmosphere system is about
E
in
=

F
s
(1 A
e.0
)
R
2
e
(9.99)
where R
e
is the Earths radius, R
2
e
is the Earths cross-sectional area (effective
area that solar irradiance impinges upon), and A
e.0
is the globally and wavelength-
averaged Earthatmosphere albedo (0.30). The outgoing energy fromthe Earth is
E
out
=
e.0
B
T
4
e
4R
2
e
(9.100)
where T
e
is the temperature of the Earths surface, and
e.0
is Earths globally
and wavelength-averaged surface emissivity. The actual average emissivity is 0.96
0.98, but if the Earth is considered a blackbody,
e.0
= 1. Equating incoming and
outgoing energy gives
T
e
=

F
s
(1 A
e.0
)
4
e.0
14
(9.101)
which is the equilibrium temperature of the Earth in the absence of a greenhouse
effect.
Example 9.12
For F
s
= 1365 W m
2
, A
e.0
= 0.3, and
e.0
= 1, (9.101) predicts T
e
= 254.8 K.
The actual average surface temperature on the Earth is about 288 K, and the
difference is due primarily to absorption by greenhouse gases.
9.8.3 Solutions to the radiative transfer equation
Equations (9.74)(9.79) can be combined to give the radiative transfer equation as
dI
.j.
dS
b
= I
.j.
(
s.
+
a.
) +
s.k.
4
2
0
1
1
I
.j
.
P
s.k..j.j
..
dj
+ F
s.
e
j
s
s.k.
4
P
s.k..j.j
s
..
s
+
a.
B
.T
(9.102)
326
The fraction of total extinction due to scattering is the single-scattering albedo,
s.
=

s.
=

s.g.
+
s.a.
+
s.h.
s.g.
+
a.g.
+
s.a.
+
a.a.
+
s.h.
+
a.h.
(9.103)
Substituting (9.103),
=
s.
+
a.
from(9.81) and d
j
s
dS
b
from(9.65)
into (9.102) gives the radiative transfer equation as
j
dI
.j.
d
= I
.j.
J
diffuse
.j.
J
direct
.j.
J
emis
.j.
(9.104)
where
J
diffuse
.j.
=
1
4
s.k.
2
0
1
1
I
.j
.
P
s.k..j.j
..
dj
(9.105)
J
direct
.j.
=
1
4
F
s.
e
j
s
s.k.
P
s.k..j.j
s
..
s
(9.106)
J
emis
.j.
= (1
s.
)B
.T
(9.107)
9.8.3.1 Analytical solutions
The radiative transfer equation must be solved numerically, except for idealized
cases. For example, an analytical solution can be derived when absorption is con-
sidered but scattering and emission are neglected. In this case, (9.104) simplies
for the upward and downward directions to
j
dI
.j.
d
a.
= I
.j.
j
dI
.j.
d
a.
= I
.j.
(9.108)
respectively, where +j is used for upward radiation, j is used for downward
radiation, and
a.
is the optical depth due to absorption only. Integrating the
upward equation from a lower optical depth
a..b
to the optical depth of interest,
a.
, gives
I
.j.
(
a.
) = I
.j.
(
a..b
)e
(
a.
a..b
)j
(9.109)
Integrating the downward equation froman upper optical depth
a..t
to the optical
depth of interest,
a.
, gives
I
.j.
(
a.
) = I
.j.
(
a..t
)e
(
a.
a..t
)j
(9.110)
Equations (9.109) and (9.110) describe Beers law, which states that the absorption
of radiation increases exponentially with the optical depth of the absorbing species.
327
A second analytical solution can be derived when absorption and infrared emis-
sion are considered, but scattering is neglected. In this case, (9.104) simplies for
the upward and downward directions as
j
dI
.j.
d
a.
= I
.j.
B
.T
j
dI
.j.
d
a.
= I
.j.
B
.T
(9.111)
respectively, which are Schwartzchilds equations. Integrating (9.111) gives upward
and downward radiances as
I
.j.
(
a.
) = I
.j.
(
a..b
)e
(
a.
a..b
)j
1
j

a.
a..b
B
.T(
a.
)
e
(
a.
a.
)j
a.
(9.112)
I
.j.
(
a.
) = I
.j.
(
a..t
)e
(
a.
a..t
)j
+
1
j

a.
a..t
B
.T(
a.
)
e
(
a.
a.
)j
a.
(9.113)
respectively. In each equation, the Planck function, which varies with temperature,
must be integrated between the two optical depths.
9.8.3.2 Numerical solutions
When particle scattering is included in the radiative transfer equation, analyti-
cal solutions become difcult to obtain and numerical solutions are needed. One
numerical solution is found with the two-stream method. With this method, radi-
ance is divided into an upward () and a downward () component, each of which
is approximated with a forward and a backward scattering term in the diffuse
phase function integral of (9.105). One approximation to the integral is
1
4
2
0
1
1
I
.j
.
P
s.k..j.j
..
dj
(1 + g
a.k.
)
2
I +
(1 g
a.k.
)
2
I upward
(1 + g
a.k.
)
2
I +
(1 g
a.k.
)
2
I downward
(9.114)
where I is the downward radiance, I is the upward radiance, and (1 + g
a.k.
)2
and (1 g
a.k.
)2 are integrated fractions of the forward- and backward-scattered
energy, respectively. Wavelength subscripts have been omitted on I and I. Equa-
tion (9.114) is the two-point quadrature approximation to the phase-function
integral (Liou 1974, 2002; Meador and Weaver 1980). Substituting (9.114) into
328
0 0.2 0.4 0.6 0.8 1
300
400
500
600
700
800
900
1000
0 0.2 0.4 0.6 0.8 1
P
r
e
s
s
u
r
e

(
h
P
a
)
With aerosols
No aerosols
g
a
w
s
0 0.2 0.4 0.6 0.8 1
300
400
500
600
700
800
900
1000
0 0.2 0.4 0.6 0.8 1
P
r
e
s
s
u
r
e

(
h
P
a
)
With aerosols
No aerosols
g
a
w
s
(a) (b)
Figure 9.31 Modeled vertical proles from Claremont, California (11.30
on August 27, 1997) of the effective asymmetry parameter and the single-
scattering albedo at (a) 0.32 and (b) 0.61 m when aerosol particles were
and were not included in the simulations. The overall asymmetry param-
eter was zero when aerosol particles were absent. The peaks at 920 hPa
indicate elevated aerosol layers. From Jacobson (1998b).
(9.105) gives
1
4
s.k.
2
0
1
1
I
.j
.
P
s.k..j.j
..
dj
s.
(1 b
)I +
s.
b
s.
(1 b
)I +
s.
b
I
(9.115)
where
s.
is the single-scattering albedo from (9.103),
1 b
=
1 + g
a.
2
b
=
1 g
a.
2
(9.116)
are effective integrated fractions of forward- and backward-scattered energy, and
g
a.
=

s.a.
g
a.a.
+
s.c.
g
a.h.
s.g.
+
s.a.
+
s.h.
(9.117)
is an effective asymmetry parameter, which is the weighted sum of asymmetry
parameters for gases (g
a.g.
), aerosol particles (g
a.a.
), and hydrometeor particles
(g
a.h.
). Asymmetry parameters for gases are zero.
Figures 9.31 (a) and (b) showvertical proles of effective asymmetry parameters
and single-scattering albedos over an urban airshed from two model simulations.
In one simulation, aerosol particles were assumed to be present, and in the other,
they were ignored. Clouds were not present in either case. The gures indicate that
effective asymmetry parameters increased from the free troposphere to the bound-
ary layer when aerosol particles were assumed to be present. Above the boundary
layer, the effective asymmetry parameters were small because Rayleigh scatter-
ing, which has a zero asymmetry parameter, dominated scattering extinction (see
329
Fig. 9.23). Within the boundary layer, effective asymmetry parameters increased
because aerosol-scattering extinction dominated Rayleigh-scattering extinction.
The single-scattering albedo curve in Fig. 9.31(a) indicates that, above the
boundary layer, ultraviolet extinction was due almost entirely to gas scattering
when aerosol particles were ignored, and to gas and particle scattering when aerosol
particles were present. Within the boundary layer, aerosol absorption played a
larger role than it did above the boundary layer.
The phase function for the single scattering of solar radiation, used in (9.106),
can be estimated with the Eddington approximation of the solar phase function
(Eddington 1916)
P
s.k.
(O
s
) 1 3g
a.
j
1
j
s
(9.118)
where j
1
is the diffusivity factor, set to j
1
= 1
3 when the quadrature approxi-

mation for diffuse radiation is used (Liou 1974).
With the parameters above, the spectral radiance from (9.104) can be written
for solar wavelengths (where the emission term is neglected) in terms of an upward
and a downward component as
j
1
dI
d
= I
s
(1 b)I
s
bI
s
4
(1 3g
a
j
1
j
s
)F
s
e
j
s
(9.119)
j
1
dI
d
= I
s
(1 b)I
s
bI

s
4
(1 +3g
a
j
1
j
s
)F
s
e
j
s
(9.120)
respectively (Liou 2002), where all wavelength subscripts have been omitted. The
term j
1
is used for direct solar radiation on the right side of (9.119) because
solar radiation is downward relative to I. The term +j
1
is used on the right side
of (9.120) for direct solar radiation since solar radiation is in the same direction
as I.
Equations (9.119) and (9.120) can be written in terms of spectral irradiance with
the conversions F= 2j
1
I and F= 2j
1
I, and generalized for the two-point
quadrature or another two-stream approximation. The resulting equations are
dF
d
=
1
F
2
F
3
s
F
s
e
j
s
dF
d
=
1
F +
2
F +(1
3
)
s
F
s
e
j
s
(9.121)
(Meador and Weaver 1980; Toon et al. 1989b; Liou 2002), where the s, dened in
Table 9.9, are coefcients derived from the two-point quadrature and Eddington
approximations of the diffuse phase-function integral. The Eddington approxi-
mation of the diffuse phase function integral is discussed in Irvine (1968, 1975),
330
Table 9.9 Coefcients for the two-stream method for two
approximations of the diffuse phase function
Approximation
1

2

3
Quadrature
1
s
(1 + g
a
)2
j
1
s
(1 g
a
)
2j
1
1 3g
a
j
1
j
s
2
Eddington
7
s
(4 +3g
a
)
4

1
s
(4 3g
a
)
4
2 3g
a
j
s
4
Sources: Liou (2002), Meador and Weaver (1980), Toon et al. (1989a).
Kawata and Irvine (1970), and Meador and Weaver (1980). Replacing the solar
term in (9.121) with an emission term gives an equation for infrared irradiance as
dF
d
=
1
F
2
F 2(1
s
)B
T
dF
d
=
1
F +
2
F +2(1
s
)B
T
(9.122)
Equations 9.1199.122 may be solved in a model after the equations are dis-
cretized, a vertical grid is dened, and boundary conditions are dened. Figure 7.1
shows a model atmosphere in which layer bottoms and tops are dened by indices,
k 0.5. For that grid, upward and downward irradiances at the bottom of a layer
k are F
k+12
and F
k+12
, respectively, where k = 0, . . . , N
L
. The derivatives in
(9.119)(9.122) can be discretized over this grid with a nite-difference expansion
of any order. A discretization requires boundary conditions at the ground and top
of the atmosphere. At the ground, boundary conditions for irradiance at a given
wavelength are
F
N
L
+12
= A
e
F
N
L
+12
+
A
e
j
s
F
s
e
N
L
+12
j
s
solar
B
T
infrared
(9.123)
where A
e
is the albedo and is the emissivity. Equation (9.123) states that the
upward irradiance at the surface equals the reected downward diffuse irradiance
plus the reected direct solar or emitted infrared irradiance. At the top of the atmo-
sphere, the boundary condition is F
12
= j
s
F
s
for solar radiation and F
12
=
F
12
for infrared radiation. The boundary conditions for radiance are similar.
Terms from the discretization of (9.119)(9.122) can be placed in a matrix. If a
second-order central difference discretization is used, the matrix is tridiagonal and
can be solved noniteratively to obtain the upward and downward uxes at the
boundary of each layer (e.g., Toon et al. 1989a).
When the atmosphere absorbs signicantly, the two-stream and Eddington
approximations underestimate forward scattering because the expansion of the
phase function is too simple to obtain a strong peak in the scattering efciency. As
a partial remedy, the effective asymmetry parameter, single-scattering albedo, and
331
optical depth can be adjusted with the delta functions
g
a
=
g
a
1 + g
a
s
=
1 g
2
a
s
1
s
g
2
a
1
s
g
2
a
(9.124)
respectively (Hansen 1969; Potter 1970; Joseph et al. 1976; Wiscombe 1977; Liou
2002), which replace terms in (9.104)(9.107) and in Table 9.9.
More advanced methods of solving the radiative transfer equation are four-
stream and higher-stream techniques (e.g., Liou 1974, 2002; Cuzzi et al. 1982; Fu
and Liou 1993; Fu et al. 1997). A four-stream approximation is one in which radi-
ance is divided into two upward and two downward components instead of one
upward and one downward component, as was done for the two-stream approxi-
mation. In addition, the phase-function integral is expanded into four terms instead
of the two terms in (9.114), and each of the four radiances is solved using one of
the four phase-function terms. The nal upward and downward radiances with
the four-stream approximation are weighted averages of the two upward and two
downward radiances, respectively.
9.8.4 Heating rates and photolysis coefcients
Spectral irradiances are used to estimate changes in air temperature due to radiative
heating and cooling. If only radiative effects are considered, the local time rate of
change of temperature of an atmospheric layer is found fromthe net ux divergence
equation,
T
t
r
=
1
c
p.m
dQ
solar
dt
+
dQ
ir
dt
=
1
c
p.m
a
F
n
z
(9.125)
where F
n
=
0
(F
)d is the net downward minus upward radiative ux

(W m
2
), summed over all wavelengths, and Q
solar
and Q
ir
are solar and infrared
radiative heating rates (J kg
1
), respectively, fromthe thermodynamic energy equa-
tion of (3.75). At the ground, F
n
= F
n.g
from (8.104). Positive values of F
n
cor-
respond to net downward radiation. The partial derivative of F
n
with respect to
altitude can be discretized over layer k with
F
n.k
z

[(F
.k12
F
.k12
) (F
.k+12
F
.k+12
)]
z
k12
z
k+12
(9.126)
where k identies the center of a layer, k 12 identies the top of a layer, and
k +12 identies the bottom of a layer. The resulting change in temperature in the
layer due to diabatic radiative heating is
LT
k
1
c
p.m
a
F
n
z
h (9.127)
where h is the time step.
332
Photolysis rate coefcients are found by solving (9.119) and (9.120) for the solar
spectral radiance. The spectral radiance is multiplied by 4 steradians to obtain
the spectral actinic ux, as dened in (9.12). The photolysis rate coefcient (s
1
) of
a species q producing product set p at the bottom of layer k (denoted by subscript
k +12) is
J
q. p.k+12
=

0
4I
p..k+12
b
a.g.q..T
Y
q. p..T
d (9.128)
where b
a.g.q..T
, previously dened, is the temperature- and wavelength-dependent
absorption cross section of gas q (cm
2
molec.
1
), Y
q. p..T
is the temperature-
and wavelength-dependent quantum yield of q producing product set p (molec.
photon
1
), and
I
p..k+12
= (I
.k+12
I
.k+12
)
10
10
m
3
cm
2
m

hc
(9.129)
(photons cm
2
s
1
m
1
sr
1
) is the spectral radiance at the bottom of layer k.
The units of I, , c, and h (Plancks constant) for use in this equation are W m
2
m
1
sr
1
, m, m s
1
, and J s, respectively. The quantum yield is the fractional
number (1) of molecules of a specic product formed per photon of radiation
absorbed at a given wavelength. A photolysis reaction may produce different sets
of products. Each set of products has its own set of wavelength-dependent quantum
yields, as discussed in Chapter 10.
Example 9.14
Convert I
= 12 W m
2
m
1
sr
1
in the wavelength region, 0.495 m - -
0.505 m, to photons cm
2
s
1
sr
1
.
SOLUTION
The solution to this problem is I
p.
= LI
10
10
hc, where the wavelength
interval is L = 0.01 m, and the mean wavelength is = 0.5 m. Solving
gives I
p.
= 3.02 10
13
photons cm
2
s
1
sr
1
.
Gases, aerosol particles, and hydrometeor particles affect climate, weather, and
air quality through their effects on heating and photolysis. For example, some stud-
ies have found that cloud drops, which are highly scattering, increase the backscat-
tered fraction of incident solar radiation, increasing photolysis within and above
a cloud and decreasing it below the cloud (e.g., Madronich 1987; van Weele and
Duynkerke 1993; de Arellano et al. 1994). Other studies have found that aerosol
particle absorption decreases photolysis below the region of absorption, decreas-
ing ozone (e.g., Jacobson 1997b, 1998b; Castro et al. 2001). Conversely, aerosol
333
particle scattering increases photolysis, potentially increasing ozone (e.g., Dicker-
son et al. 1997; Jacobson 1998b).
9.9 SUMMARY
In this chapter, radiation laws, radiation processes, optical properties, and the radia-
tive transfer equation were described. The laws include Plancks law, Wiens law,
and the StefanBoltzmann law. The radiation processes include reection, refrac-
tion, scattering, absorption, diffraction, and transmission. Gases, aerosol parti-
cles, and hydrometeor particles attenuate radiation by absorbing and scattering.
All gases scatter ultraviolet and short visible wavelengths. Some gases selectively
absorb ultraviolet, visible, and infrared wavelengths. All aerosol particle compo-
nents scatter visible light, but only a few absorb such radiation. Hydrometeor
particles almost exclusively scatter visible light, except at large size. For example,
large raindrops absorb visible light. The radiative transfer equation determines the
change in radiance and irradiance along a beam due to scattering out of the beam,
absorption along the beam, multiple scattering of diffuse radiation into the beam,
single scattering of direct solar radiation into the beam, and emission of infrared
radiation into the beam. Changes in radiance and irradiance are used to calculate
photolysis and heating rates, respectively.
9.10 PROBLEMS
9.1 Calculate the radiance and irradiance from the Planck function at T =
273 K and (a) = 0.4 m, (b) = 1.0 m, and (c) = 15 m.
9.2 Calculate the equilibrium surface temperature of the Earth if its emissivity
and albedo were 70 and 25 percent, respectively. Does this represent an
increase or decrease with respect to the equilibrium temperature (when the
emissivity and albedo are 100 and 30 percent, respectively)?
9.3 Calculate the transmission of light through a 0.5-m-diameter particle made
of ammonium sulfate at a wavelength of = 0.5 m (use the data from
Table 9.6).
9.4 Calculate the extinction coefcient, meteorological range, and optical depth
in a 1-km region of the atmosphere resulting from nitrogen dioxide absorp-
tion at a wavelength of = 0.55 m when the volume mixing ratio of NO
2
is = 0.05 ppmv, T = 288 K, and p
d
= 980 hPa.
9.5 Calculate the real and imaginary refractive indices of a two-component mix-
ture using (a) the volume average mixing rule and (b) the volume average
dielectric constant mixing rule when the volume fraction of each component
is 0.5 and the index of refraction of the matrix component is m
1
= 1.34
i1.0 10
9
and that of the absorbing inclusion is m
2
= 1.82 i0.74. Also,
calculate the real and complex parts of the dielectric constant for each of the
two components in the mixture.
9.6 Find the meteorological ranges at = 0.53 m under the following
conditions:
334
(a) In a rain shower with 1-mm-diameter water drops and a mass loading
of 1 g m
3
.
(b) In a fog with 10-m-diameter water drops and a mass loading of 1 g m
3
.
(c) In a haze with 0.5-m-diameter particles, 40 percent ammonium sulfate
by volume, 60 percent liquid water by volume, and a total mass loading
of 50 g m
3
.
(d) Behind a diesel exhaust, with 50 g m
3
of 0.5-m-diameter soot
particles.
Inall cases, assume particles are spherical, neglect Rayleighscattering,
and neglect gas absorption. In the case of soot and raindrops, assume
particle scattering and absorption occur. In the other cases, assume only
particle scattering occurs. In the case of haze, calculate the scattering
efciency by weighting the volume fractions of each component by the
corresponding scattering efciency. Use Figs. 9.20 and 9.21 to determine
the efciencies. Assume the scattering efciency of ammoniumsulfate is
the same as that of water. Assume the mass densities of soot, ammonium
sulfate, and liquid water are 1.25, 1.77, and 1.0 g cm
3
, respectively.
9.7 Write a script to calculate radiance and irradiance from the Planck function
versus temperature and wavelength. Use the program to calculate values
between = 0.01 m and = 100 mm for T = 6000 K and T = 300 K. Plot
the results.
9.8 Divide the atmosphere from z = 0 to 10 km into 100 vertical layers. Assume
T = 288 K and p
d
= 1013 hPa at the surface, and assume the temperature
decreases 6.5 K km
1
. Use (2.41) to estimate the air pressure in each layer.
Calculate the extinction coefcient and optical depth in each layer due to
Rayleigh scattering at = 0.4 m. Estimate the cumulative optical depth at
the surface.
335
10
Gas-phase species, chemical reactions,
and reaction rates
T
he atmosphere contains numerous gases that undergo chemical reaction.
Because many chemical pathways are initiated by sunlight, atmospheric reac-
tions are collectively called photochemical reactions. Lightning, changes in temper-
ature, and molecular collisions, though, also initiate reactions. Photochemistry is
responsible for the transformation of gases in all regions of the atmosphere. It con-
verts nitrogen oxide and reactive organic gases, emitted during fuel combustion, to
ozone, peroxyacetyl nitrate (PAN), and other products. It produces and destroys
ozone in the free troposphere and stratosphere. Photochemistry also converts
dimethyl sulde (DMS) to sulfuric acid over the oceans. In this chapter, chemical
species, structures, reactions, rate coefcients, rates, and lifetimes are discussed. In
Chapter 11, photochemical reactions important in different regions of the atmo-
sphere are described. In Chapter 12, numerical methods of solving chemical ordi-
nary differential equations arising from such reactions are given.
10.1 ATMOSPHERIC GASES AND THEIR
MOLECULAR STRUCTURES
Gases consist of neutral or charged single atoms or molecules. Elements that make
up most gases are hydrogen (H), carbon (C), nitrogen (N), oxygen (O), uorine
(F), sulfur (S), chlorine (Cl), and bromine (Br). Particles contain sodium (Na),
magnesium (Mg), aluminum (Al), silicon (Si), potassium (K), calcium (Ca), and/or
the elements found in gases.
Gas molecules consist of atoms with covalent bonds between them. A covalent
bond is a bond consisting of one or more pairs of electrons shared between two
atoms. Gilbert Lewis (18751946) suggested that atoms could be held together by
shared electron pairs. Such sharing occurs in the outer shell (valence shell) of an
atom. Lewis proposed the use of electron-dot symbols (Lewis symbols) to describe
the conguration of electrons in the valence shell of an atom. Lewis symbols are
drawn by arranging electrons around an element as dots. Asingle dot represents an
unpaired electron and two dots adjacent to each other represent an electron pair.
Table 10.1 shows Lewis symbols for several elements in the periodic table.
Hydrogen (H), the rst element of the periodic table, has one electron in its
valence shell, as shown by its Lewis symbol in Table 10.1. The electron in H is
unpaired. The number of unpaired electrons in the valence shell of an element
is the valence of the element. If a hydrogen atom covalently bonds with another
336
10.1 Atmospheric gases and their molecular structures
Table 10.1 Elements in periods 14 and groups IVIII of the periodic table and
their Lewis symbols.
Group

Period
I II III IV V VI VII VIII
1
1 1
H

1.008
2 0
He

4.003
2
3 1
Li

6.941
4 2
Be

9.012
5 3
B

10.81
6 4
C

12.01
7 3
N

14.01
8 2
O

16.00
9 1
F

19.00
10 0
Ne

20.18
3
11 1
Na

22.99
12 2
Mg

24.30
13 3
Al

26.98
14 4
Si

28.09
15 3
P

30.97
16 2
S

32.07
17 1
Cl

35.45
18 0
Ar

39.95
4
19 1
K

39.10
20 2
Ca

40.08
31 3
Ga

69.72
32 4
Ge

72.61
33 3
As

74.92
34 2
Se

78.96
35 1
Br

79.90
36 0
Kr

83.80
Note: The top left number in each entry is the atomic number. The top right number is the
principal valence of the element. The dots are the Lewis symbols of the element. The bottom
number is the atomic weight of the element (g mol
1
). The group number is also the number of
valence-shell electrons of each element, except for helium, which has two valence-shell electrons.
hydrogen atom, each atom contributes one electron to the valence shell of the
other, giving each atom two electrons (an electron pair) in its valence shell. The
resulting molecule in this case is H
:
H (H
2
, molecular hydrogen), where the two
dots indicate an electron pair. The maximum number of electrons in the valence
shell of a hydrogen atom is two, so a hydrogen atom can bond with a maximum of
one other hydrogen atom (or another atom). The maximum number of hydrogen
atoms that can bond with one atom of an element before the valence shell of the
element is full is called the principal valence of the element. The principal valence
of hydrogen is one (Table 10.1).
Helium (He), the second element of the periodic table, has two electrons in
its outer shell (Table 10.1), and they are paired together. Since heliums electrons
are paired, helium has a valence of zero, and it cannot covalently bond with any
hydrogen atoms, so heliumis chemically inert (unreactive), and its principal valence
is zero.
Whereas hydrogen and helium, which are in period (row) 1 of the periodic table,
can have a maximum of two electrons in their outer shell, elements in periods 2,
3, and 4 can hold up to eight electrons in their outer shell. In these three periods,
elements in groups I, II, III, and IV(Table 10.1) have 1, 2, 3, or 4 unpaired electrons,
respectively, in their outer shell. Thus, the elements in groups I, II, III, and IV can
bond with 1, 2, 3, or 4 hydrogen atoms, respectively, resulting in 2, 4, 6, or 8
electrons, respectively, in their outer shells.
337
Gas-phase species, chemical reactions, and reaction rates
Elements in groups V, VI, VII, and VIII (Table 10.1) have 1 paired and 3 unpaired
electrons, 2 paired and 2 unpaired electrons, 3 paired and 1 unpaired electrons,
and 4 paired and 0 unpaired electrons, respectively, in their outer shells. Thus, each
element can bond with 3, 2, 1, or 0 hydrogen atoms, respectively, resulting in eight
outer-shell electrons in all cases.
Avalence shell with eight electrons is called an octet. This is the most stable form
of a valence shell. Atoms that have their valence shells lled with eight electrons
(Ne, Ar, Kr) are so stable, they are chemically inert. Atoms with unpaired electrons
in their valence shells try to share electrons with other atoms (through chemical
reaction) to become more stable, so they are chemically reactive.
When atoms form covalent bonds during production of a molecule, the cong-
uration of the molecule can be illustrated with a Lewis structure, which consists
of a combination of Lewis symbols. Appendix Table B.3 shows the Lewis struc-
tures of many gases of atmospheric importance. The table is divided into inorganic
and organic gases. Inorganic gases are those that contain any element, including
hydrogen (H) or carbon (C), but not both H and C. Organic gases are those that
contain both Hand C, but may also contain other elements. Organic gases that con-
tain only H and C are hydrocarbons. Hydrocarbons include alkanes, cycloalkanes,
alkenes, alkynes, aromatics, and terpenes. When methane, which is fairly unre-
active, is excluded from the list of hydrocarbons, the remaining hydrocarbons are
nonmethane hydrocarbons (NMHCs). Oxygenated functional groups, such as alde-
hydes, ketones, alcohols, acids, and nitrates, are added to hydrocarbons to produce
oxygenated hydrocarbons. Nonmethane hydrocarbons and oxygenated hydrocar-
bons, together, are reactive organic gases (ROGs) or volatile organic carbon (VOC).
Nonmethane hydrocarbons and carbonyls (aldehydes plus ketones), together, are
nonmethane organic carbon (NMOC). Total organic gas (TOG) is the sum of
ROGs and methane. Below, the Lewis structures and characteristics of a few inor-
ganic and organic gases are discussed.
10.1.1 Molecular hydrogen (H
2
)
Molecular hydrogen is emitted fromvolcanos, produced biologically, and produced
chemically in the atmosphere. It has a typical mixing ratio of about 0.6 ppmv in
the troposphere and consists of two hydrogen atoms with a single covalent bond
between them. Two Lewis structures for molecular hydrogen are
H
:
H H
H
The two dots and the line between atoms indicate a covalently bonding electron
pair.
10.1.2 Molecular oxygen (O
2
)
Molecular oxygen is produced by green-plant photosynthesis and has a mixing
ratio of 20.95 percent by volume throughout the homosphere. Its Lewis structures
are
:
..
O
::
..
O
:
O
=
O
338
where the double bond indicates two shared electron pairs. In the second structure,
all unshared pairs of electrons are ignored for convenience.
10.1.3 Molecular nitrogen (N
2
)
Sources of molecular nitrogen were discussed in Chapter 2. Its mixing ratio is
typically 78.08 percent by volume throughout the homosphere. Molecular nitrogen
has the Lewis structures,
:
N
:::
N
:
N
N
where the triple bond indicates three shared electron pairs.
10.1.4 Hydroxyl radical (OH)
Molecular hydrogen and molecular oxygen have no unpaired electrons. Molecules
that have one unpaired electron are free radicals. One free radical is the hydroxyl
radical, which has the Lewis structures,
:
..
O
:
H

O
H
The single dot in the second structure indicates that the unpaired electron is asso-
ciated with the oxygen atom. The hydroxyl radical is produced chemically in the
atmosphere. It breaks down many other gases and is referred to as a scavenger.
OH is discussed in more detail in Chapter 11.
10.1.5 Nitric oxide (NO)
Nitric oxide is also a free radical. It is a colorless gas emitted from soils, plants,
and combustion processes and produced by lightning and chemical reaction. Com-
bustion sources include aircraft, automobiles, oil reneries, and biomass burning.
The primary sink of NO is chemical reaction. A typical mixing ratio of NO in
the background troposphere near sea level is 5 pptv. In the upper troposphere NO
mixing ratios increase to 2060 pptv. In urban regions, NOmixing ratios reach 0.1
ppmv in the early morning but decrease signicantly by midmorning. The Lewis
structures of NO can be drawn so that either N or O has seven electrons in its
valence shell. Thus,
:
N
::
..
O
:

:
..
N
::
O
:
+

N
=
O

N
=
O
+
where each structure is a resonance structure. A resonance structure is one of two
or more Lewis structures used to represent a molecule that cannot be represented
correctly by one Lewis structure. Multiple Lewis structures arise because electrons
are not local to one atom. The charge distribution shown on the second Lewis
structure of each pair indicates that the molecule may be polar (charged oppositely
on either end). Since total charges balance, there is no net charge on the molecule
as a whole.
339
10.1.6 Nitrogen dioxide (NO
2
)
Nitrogen dioxide is a brown gas because it absorbs the shortest wavelengths of
the Suns visible radiation. It absorbs almost all blue light and some green light,
allowing the remaining green light and all red light to scatter and transmit. The
combination of red and some green light is brown. The major source of atmospheric
NO
2
is photochemical oxidation of NO. NO
2
is also produced by other reaction
pathways and is emitted during combustion. NO
2
is more prevalent during mid-
morning than during midday or afternoon, since sunlight breaks NO
2
down past
midmorning. Mixing ratios of NO
2
just above sea level in the free troposphere
range from 20 to 50 pptv. In the upper troposphere, mixing ratios increase to 30
70 pptv. In urban regions, they range from 0.1 to 0.25 ppmv. The resonance struc-
tures of NO
2
are
N
O
O
N
O O
+
+

N
O O
+
N
O O
+

During nitrogen dioxide formation, a net negative charge is transferred to the
oxygen atoms from the nitrogen atom, resulting in the charge distribution shown.
10.1.7 Ozone (O
3
)
Ozone is a colorless gas that exhibits an odor, even in small concentrations. In
urban areas, ozone affects human health in the short term by causing headache
(>0.15 ppmv), chest pain (>0.25 ppmv), and sore throat and cough (>0.30 ppmv).
Ozone decreases lung function for those who exercise steadily for over an hour
while exposed to concentrations above 0.30 ppmv. Above 1 ppmv (high above
ambient concentrations), ozone can temporarily narrow passages deep in the lung,
increasing airway resistance and inhibiting breathing. Small decreases in lung func-
tion affect those with asthma, chronic bronchitis, and emphysema. Ozone may
also accelerate the aging of lung tissue. Above 0.1 ppmv, ozone affects animals
by increasing their susceptibility to bacterial infection. It also interferes with the
growth of plants and trees and deteriorates organic materials, such as rubber, textile
dyes and bers, and some paints and coatings (USEPA 1978).
In the free troposphere, ozone mixing ratios range from 20 to 40 ppbv near sea
level and from 30 to 70 ppbv at higher altitudes. In urban areas, ozone mixing
ratios range from 0.01 ppmv at night to 0.35 ppmv during smoggy afternoons,
with typical values of 0.15 ppmv during moderately polluted afternoons. Ozone is
chemically reactive but not a free radical. The resonance structures of ozone are
O
O O
O
O O
O
O
O
O
O O
+ +
+
+

340
10.1.8 Carbon monoxide (CO)
Carbon monoxide is a tasteless, colorless, and odorless gas that is toxic to humans
and animals exposed to it for one hour at mixing ratios above about 700 ppmv.
Exposure to 300 ppmv for one hour causes headaches. In urban regions away from
freeways, CO mixing ratios are typically 210 ppmv. On freeways and in trafc
tunnels, mixing ratios can rise to more than 100 ppmv. In the free troposphere, CO
mixing ratios vary from 50 to 150 ppbv. A major source of CO is incomplete com-
bustion by automobiles, trucks, and airplanes. Wood burning and grass burning
are also important sources. Natural sources of COare plants and biological activity
in the oceans. Although CO is the most abundantly emitted pollutant gas in urban
air, it does not play a major role in photochemical smog formation. Photochemical
smog is characterized by the buildup of ozone and related products. CO does not
produce much ozone in urban air. The major sink of CO is chemical conversion to
carbon dioxide (CO
2
). CO can be represented by
:
C
:::
O
:
+
C
O
+
10.1.9 Carbon dioxide (CO
2
)
Carbon dioxide is an odorless and inert gas. Its major sources and sinks were
described in Chapter 2. It is well mixed throughout the troposphere and strato-
sphere, with a current mixing ratio of about 375 ppmv. The Lewis structures of
CO
2
are
C O O
O C O
These structures indicate that CO
2
is a linear molecule but not a free radical. CO
2
is very stable: its lifetime against chemical destruction is over 100 years.
10.1.10 Sulfur dioxide (SO
2
)
Sulfur dioxide is a colorless gas that exhibits an odor and taste at high concentra-
tions. It is emitted fromcoal-red power plants, automobile tailpipes, and volcanos.
It is also produced chemically in the atmosphere from biologically emitted precur-
sors, such as dimethylsulde (DMS) and hydrogen sulde (H
2
S). In the background
troposphere, SO
2
mixing ratios range from 20 pptv to 1 ppbv. In moderately pol-
luted air, they range from 10 to 30 ppbv. SO
2
is removed from the atmosphere by
gas-phase reaction, dissolution into clouds and rain, and deposition to the ground.
Common Lewis structures of SO
2
are
S
O
O
S
O O
341
10.1.11 Methane (CH
4
)
Sources and sinks of methane were discussed in Chapter 2. It is the most unreac-
tive hydrocarbon in the atmosphere. Because methane has a long lifetime against
chemical loss (about 10 years), it is well diluted throughout the free troposphere,
with a mixing ratio of about 1.8 ppmv. The Lewis structures of methane are
C H H
H
H
H C
H
H
H
10.1.12 Peroxyacetyl nitrate (PAN) (CH
3
C(O)OONO
2
)
PAN is an eye irritant, initially discovered in the laboratory as a product of smog-
forming chemical reactions. Its peak mixing ratio of about 1020 ppbv in polluted
air occurs at the same time during the afternoon as does ozones peak mixing ratio.
PAN mixing ratio in the free troposphere ranges from 2 to 100 pptv. PAN can be
represented by
C C H
H
H
O
O
O
N
O
O
+
C C
O
O
H
H
H
O
N
O
O
+
10.2 CHEMICAL REACTIONS AND PHOTOPROCESSES
A single chemical reaction with no intermediate products is an elementary reac-
tion. Elementary homogeneous gas-phase chemical reactions in the atmosphere are
conveniently divided into photolysis reactions (also called photoprocesses, pho-
todissociation reactions, or photolytic reactions) and chemical kinetic reactions.
Photolysis reactions are unimolecular (one-body) reactions initiated when a pho-
ton of radiation strikes a molecule and breaks it into two or more products. Ele-
mentary chemical kinetic reactions are bimolecular (two-body) or termolecular
(three-body). Reactants and products of photolysis and kinetic reactions are neu-
tral or charged atoms or molecules.
Photolysis reactions are unimolecular. An example of a photolysis reaction is
NO
2
+h
NO+
- 420 nm (10.1)
where h is a single photon of radiation, O [= O(
3
P)] is ground-state atomic oxy-
gen, and paired electrons and bonds between atoms are not shown (only unpaired
electrons are shown).
Elementary bimolecular reactions include thermal decomposition, isomeriza-
tion, and standard collision reactions. Thermal decomposition and isomerization
reactions occur when a reactant molecule collides with an air molecule. The kinetic
342
10.2 Chemical reactions and photoprocesses
energy of the collision elevates the reactant to a high enough vibrational energy
state that it can decompose or isomerize. Thermal decomposition occurs when the
excited reactant dissociates into two or more products. Isomerization occurs when
the excited reactant changes chemical structure but not composition or molecular
weight.
An example of a bimolecular thermal decomposition reaction is
N
2
O
5
+M
NO
2
+N
O
3
+M (10.2)
where N
2
O
5
is dinitrogen pentoxide, NO
3
is the nitrate radical, and M is the
molecule that provides the collisional energy. In the atmosphere, M can be any
molecule. Because oxygen and nitrogen, together, make up more than 99 percent
of the gas molecules in the atmosphere, the molecule that M represents is usually
oxygen or nitrogen. For rate calculation purposes, the concentration of Mis usually
set equal to the concentration of total air (Equation 10.20).
Since M in (10.2) does not change concentration, the reaction can be written in
nonelementary form as
N
2
O
5
M
NO
2
+N
O
3
(10.3)
Thermal decomposition reactions are temperature dependent. At high tempera-
tures, they proceed faster than at low temperatures. Isomerization reactions are
similar to (10.2) and (10.3), except that an isomerization reaction has one prod-
uct, which is another form of the reactant.
Elementary bimolecular collision reactions are the most common types of kinetic
reaction and may occur between any two chemically active reactants that collide.
A prototypical collision reaction is
CH
4
+
OH
CH
3
+H
2
O (10.4)
where CH
3
is the methyl radical, and H
2
O is water vapor. In some cases, bimolec-
ular reactions result in collision complexes that ultimately break into products.
Such reactions have the form A +B
AB
D+F, where AB
is a molecule that
has weak bonds and stays intact slightly longer than the characteristic time of
the molecules vibrations and rotations. Other reactions that may form collision
complexes are pressure-dependent termolecular reactions.
Termolecular reactions often consist of pairs of elementary bimolecular reac-
tions. Consider the termolecular combination reaction
NO
2
+N
O
3
+MN
2
O
5
+M (10.5)
The sequence of elementary bimolecular reactions resulting in (10.5) is A +B
AB
followed by AB
+MAB +M, where M is a third body whose purpose

is to carry away energy released during the second reaction. In the absence of M,
the energy release causes AB
to dissociate back to A and B. The purpose of M in

(10.5) differs from that in (10.2), which was to provide collisional energy for the
reaction. In both cases, the concentration of M is taken as that of total air for rate
calculation purposes. Reactions (10.2) and (10.5) are pressure-dependent reactions
343
because the concentration of M is proportional to air pressure. Since M in (10.5)
does not change concentration, (10.5) can also be written as
NO
2
+N
O
3
M
N
2
O
5
(10.6)
Elementary termolecular collision reactions are rare, since the probability that
three trace gases collide and change form is not large. One possible reaction of this
type is
NO+
NO+O
2

NO
2
+
NO
2
(10.7)
10.3 REACTION RATES
A reaction rate is the time rate of change of concentration of any reactant in
a reaction. The rate of an elementary photolysis, collision, isomerization, ther-
mal decomposition, or combination reaction equals a rate coefcient multiplied
by the concentration of each reactant. A rate coefcient relates concentrations
to a reaction rate. Elementary unimolecular reactions have rst-order rate coef-
cients, meaning that such coefcients are multiplied by one reactant concentration.
Elementary bimolecular and termolecular reactions have second- and third-order
rate coefcients, respectively. If reactant concentrations are expressed in units of
molecules of gas per cubic centimeter of air, the rate of reaction is in units of molec.
cm
3
s
1
, regardless of whether the reaction has a rst-, second-, or third-order
rate coefcient. Rate expressions for reactions with rst-, second-, and third-order
rate coefcients are
Rate = k
F
[A] Rate = k
S
[A][B] Rate = k
T
[A][B][C] (10.8)
respectively, where brackets denote number concentration, and k
F
. k
S
. andk
T
are
rst-, second-, and third-order rate coefcients, in units of s
1
, cm
3
molec.
1
s
1
,
and cm
6
molec.
2
s
1
, respectively. For a photolysis reaction, the rate expression is
Rate = J [A] (10.9)
where J is a rst-order photolysis rate coefcient of species A (s
1
).
If the concentration of one reactant, such as [M], is invariant during a reaction,
the concentration can be premultiplied by the rate coefcient. In such cases, second-
order rate coefcients become pseudo-rst-order coefcients and third-order coef-
cients become pseudo-second-order coefcients.
The time rate of change of concentration of a reactant equals the negative of its
rate of reaction. For a photolysis reaction of the form A +h D+G, the loss
rate of A is
d[A]
dt
= Rate = J [A] (10.10)
344
10.3 Reaction rates
For a bimolecular thermal decomposition reaction of the form, A
M
D+E, the
loss rate of A is
d[A]
dt
= Rate = k
F
[A] (10.11)
where k
F
= k
S
[M] is a pseudo-rst-order rate coefcient. For a bimolecular collision
reaction of the form A +B D+F or a termolecular combination reaction of the
form A +B
M
E, the loss rates of A and B are
d[A]
dt
=
d[B]
dt
= Rate = k
S
[A][B] (10.12)
where k
S
is a second-order rate coefcient for the rst reaction, and k
S
= k
T
[M] is
a pseudo-second-order rate coefcient for the second reaction. For a bimolecular
collision reaction of the form A +A E +F, the loss rate of A is
d[A]
dt
= 2Rate = 2k
S
[A]
2
(10.13)
where k
S
is a second-order rate coefcient. For a termolecular reaction, such as
A +B +C E +F, the loss rate of A, B, and C is
d[A]
dt
=
d[B]
dt
=
d[C]
dt
= Rate = k
T
[A][B][C] (10.14)
For a termolecular reaction such as 2A +B E +F,
d[A]
dt
= 2
d[B]
dt
= 2Rate = 2k
T
[A]
2
[B] (10.15)
In general, reactions of the form aA +bB eE +fF have the rate expression
Rate = k
r
[A]
a
[B]
b
(10.16)
where k
r
denotes a rate coefcient of order a +b. The rates of loss of species A and
B are
d[A]
dt
= aRate = ak
r
[A]
a
[B]
b
d[B]
dt
= bRate = bk
r
[A]
a
[B]
b
(10.17)
respectively, and the rates of production of E and F are
d[E]
dt
= eRate = ek
r
[A]
a
[B]
b
d[F]
dt
= fRate = fk
r
[A]
a
[B]
b
(10.18)
respectively. When M is treated as a reactant and product, such as in the reaction,
A +B +ME +M, the rate of change of M is zero. Thus,
d[M]
dt
= k
T
[A][B][M] k
T
[A][B][M] = 0 (10.19)
The rate of production of E and the rates of loss of A and B are nonzero in the
same reaction.
345
The concentration (molecules per cubic centimeter of air) of M is that of total
air,
[M] = N
a
=
p
a
k
B
T
(10.20)
where p
a
is total air pressure (hPa), k
B
19
cm
3
hPa K
1
molec.
1
), and T is absolute temperature (K). Total air concentration
and total air pressure are the sums of the concentrations and partial pressures,
respectively, of dry air and water vapor. Thus, N
a
= N
d
+ N
v
and p
a
= p
d
+ p
v
.
The concentration (molec. cm
3
) of molecular nitrogen and molecular oxygen
in the air, often needed to calculate reaction rates, are
[N
2
] =
N
2
N
d
[O
2
] =
O
2
N
d
(10.21)
where
N
2
= 0.7808 and
O
2
= 0.2095 are the volume mixing ratios of molecular
nitrogen and oxygen, respectively, from Table 2.1.
Example 10.1
Find the number concentrations of M (as total air), N
2
, O
2
, when T = 278 K
and p
a
= 920 hPa and the air is dry.
SOLUTION
From (10.20), the concentration of total air is [M] = 2.40 10
19
molec. cm
3
.
From (10.21), [N
2
] = 1.87 10
19
molec. cm
3
and [O
2
] = 5.02 10
18
molec.
cm
3
.
10.4 REACTION RATE COEFFICIENTS
10.4.1 Determining rate coefcients
Rate coefcients of kinetic reactions, which vary with temperature and pressure,
are determined experimentally. A common method of determining coefcients for
elementary bimolecular reactions at a given temperature and pressure is with kinetic
analysis. With this method, a small amount of one substance is exposed to a large
amount of another substance, and the rate of decay of the less abundant substance is
measured. Consider the bimolecular reaction A +B D+F. If a small quantity
of A is exposed to a large quantity of B, the maximum loss of B is [A]. Since
[A] [B], [B] can be held constant, and the rate of loss of A is approximately
d[A]
t
dt
= k
F
[A]
t
= k
S
[A]
t
[B]
0
(10.22)
where the subscript t indicates that the concentration changes with time, the sub-
script 0 indicates that the concentration is xed to its initial value, and k
F
= k
S
[B]
0
346
10.4 Reaction rate coefficients
is a pseudo-rst-order rate coefcient. Integrating (10.22) from time t = 0 to h and
from [A]
0
to [A]
h
and solving gives
k
S
=
1
[B]
0
h
ln
[A]
t=h
[A]
0
(10.23)
If the concentration of A at any time t = h is measured, the rate coefcient of
the reaction can be calculated from (10.23). For elementary uni- and termolecu-
lar reactions, similar calculations can be performed. For example, for unimolec-
ular reactions of the form A D+E and termolecular reactions of the form
A +B +C E +F, rate coefcients can be calculated with
k
F
=
1
h
ln
[A]
t=h
[A]
0
k
T
=
1
[B]
0
[C]
0
h
ln
[A]
t=h
[A]
0
(10.24)
respectively. In the latter case, [A]
0
[B]
0
and [A]
0
[C]
0
. More advanced meth-
ods of calculating reaction rate coefcients include fast owsystems, ash photoly-
sis, static reaction systems, pulse radiolysis, the cavity ringdown method, and static
techniques (e.g., Finlayson-Pitts and Pitts 2000).
10.4.2 Temperature dependence of reactions
First-, second-, and third-order reaction rate coefcients vary with temperature. In
many cases, the temperature dependence is estimated with the equation proposed
by Svante Arrhenius in 1889. Arrhenius is also known as the person who rst
proposed the theory of global warming, in 1896. The Arrhenius equation is found
by integrating
d(lnk
r
)
dT
=
E
r
R
T
2
(10.25)
where k
r
is the rate coefcient, E
r
is the activation energy of the reaction (J mol
1
),
T is temperature (K), and R
is the universal gas constant.

The activation energy is the smallest amount of energy required for reacting
species to form an activated complex or transition state before forming products.
Activation energies are determined by rst integrating (10.25) as
lnk
r
= ln A
r
E
r
R
T
(10.26)
where A
r
is a constant of integration, called the collisional prefactor (frequency
factor). The collisional prefactor is proportional to the frequency of those colli-
sions with proper orientation for producing a reaction. It equals a collision fre-
quency multiplied by a steric (efciency) factor. The collision frequency depends
on the relative size, charge, kinetic energies, and molecular weights of the reactant
molecules. The steric factor gives the fraction of collisions that result in an effective
347
1/T
l
n

k
r
ln A
r
Figure 10.1 Plot of ln k
r
versus 1T
(thick line). The slope of the line is
E
r
R
. The line can be extrapolated to

1T 0 (T ) to obtain ln A
r
.
reaction. For rst- and second-order reactions, collisional prefactors and rate coef-
cients have units of s
1
and cm
3
molec.
1
s
1
, respectively.
The activation energy and collisional prefactor in (10.26) are found experimen-
tally. If the rate coefcient (k
r
) is measured at different temperatures, a graph of ln
k
r
versus 1T can be plotted. E
r
and A
r
are then extracted from the plot, as shown
in Fig. 10.1.
Solving (10.26) for k
r
gives
k
r
= A
r
exp
E
r
R
= A
r
exp
C
r
T
(10.27)
where exp(E
r
R
T) is the fraction of reactant molecules having the critical energy,

E
r
, required for the reaction to occur, and C
r
= E
r
R
.
The collisional prefactor is usually a weak function of temperature, but when
the activation energy is near zero, it is a strong function of temperature. In such
cases, a temperature factor B
r
is added to (10.27), so that
k
r
= A
r
300
T
B
r
exp
C
r
T
(10.28)
B
r
is found by tting the expression in (10.28) to data. Many combination reactions
have the formof (10.28). Rate coefcients for such reactions are given in Appendix
Table B.4.
Example 10.2
At T = 298 K, the rate coefcients (cm
3
molec.
1
s
1
and cm
6
molec.
2
s
1
)
for a bi- and a termolecular reaction are, respectively,
NO+O
3

NO
2
+O
2
k
1
= 1.80 10
12
exp(1370T)
= 1.81 10
14
+O
2
+MO
3
+M k
2
= 6.00 10
34
(300T)
2.3
= 6.09 10
34
348
10.4 Reaction rate coefficients
10.4.3 Pressure dependence of reactions
Thermal decomposition and combination reactions, both of which include an M,
are pressure dependent because [M] varies with pressure. Such reactions have the
forms
A
M
D+E A +B
M
E (10.29)
respectively. Overall rate coefcients for these reactions may be interpolated
between a low- and a high-pressure limit rate coefcient. In some reactions, such
as O+O
2
+MO
3
+M, the rate coefcient is not interpolated.
The interpolation formula for the overall rate coefcient of a pressure-dependent
reaction is
k
r
=
k
.T
k
0.T
[M]
k
.T
+k
0.T
[M]
F
1+
log
10
k
0.T
[M]
k
.T
1
c
(10.30)
(Troe 1979) where F
c
is called the broadening factor of the falloff curve and
is determined theoretically, k
0.T
is the low-pressure limit rate coefcient, and
k
.T
is the high-pressure limit rate coefcient. For the reactions in (10.29),
k
r
has units of s
1
and cm
3
molec.
1
s
1
, respectively. Values of k
0.T
, k
.T
,
and F
c
are given in Table B.4 of Appendix B for applicable pressure-dependent
reactions.
The low-pressure limit rate coefcient k
0.T
, multiplied by [M], is a laboratory-
determined coefcient of the overall reaction as [M] approaches zero. Thus,
k
0.T
[M] = lim
[M]0
k
r
(10.31)
where [M] has units of molec. cm
3
, and k
0.T
has units of cm
3
molec.
1
s
1
or cm
6
molec.
2
s
1
, respectively, for the reactions in (10.29).
The high-pressure limit rate coefcient k
.T
is the rate of the overall reaction at
innite M concentration. Thus,
k
.T
= lim
[M]
k
r
(10.32)
Since the high-pressure limit rate coefcients are independent of [M], the reactions
corresponding to (10.29) at high pressure are
A D+E A +B E (10.33)
respectively. The high-pressure limit rate coefcients have units of s
1
and cm
3
molecule
1
s
1
for these two reactions, respectively.
349
10
6
10
5
10
4
10
3
1 10 100 1000
P
A
N

r
a
t
e

c
o
e
f
f
i
c
i
e
n
t

(
s
1
)
Pressure (hPa)
298 K
275 K
(a)
10
17
10
15
10
13
10
11
10
9
10
7
10
5
10
3
10
1
200 250 300
P
A
N

r
a
t
e

c
o
e
f
f
i
c
i
e
n
t

(
s
1
)
Temperature (K)
1013 hPa
(b)
Figure 10.2 Rate coefcient of PAN decomposition (a) as a function of pres-
sure for two temperatures and (b) as a function of temperature for one
pressure.
Example 10.3
When p
a
= 140 hPa, T = 216 K (stratosphere), and the air is dry, nd the rate
coefcient of
OH+
NO
2
M
HNO
3
(10.34)
SOLUTION
From(10.20), [M] =4.6910
18
molec. cm
3
. FromAppendix Table B.4, k
0.T
=
2.60 10
30
(300T)
2.9
cm
6
molec.
2
s
1
, k
.T
= 6.70 10
11
(300T)
0.6
cm
3
molec.
1
s
1
, and F
c
= 0.43. Thus, at 216 K, k
0.T
[M] = 3.16 10
11
, k
.T
=
8.16 10
11
, and k
r
= 1.11 10
11
cm
3
molec.
1
s
1
.
Figures 10.2 (a) and (b) show the pressure and temperature dependences of the
rate of PAN decomposition by the reaction
Peroxyacetyl
radical
C C
O
O
H
H
H
O
+
NO
2
Peroxyacetyl nitrate
C C
O
O
H
H
H
O
N
O
O
M
+
(10.34)
The graphs indicate that the pseudo-rst-order rate coefcient for PAN decompo-
sition varies by an order of magnitude when pressure ranges from 1 to 1000 hPa.
The rate coefcient varies by 15 orders of magnitude when temperatures range
from 180 to 330 K. Temperature affects PAN thermal decomposition much more
than does pressure.
350
10.5 Sets of reactions
10.4.4 Photolysis reactions
Photolysis reactions initiate many atmospheric chemical pathways. From (9.128),
the photolysis coefcient (s
1
) of gas q producing product set p was
J
q. p
=

0
4I
p.
b
a.g.q..T
Y
q. p..T
d (10.35)
where 4I
p.
is the actinic ux (photons cm
2
m
1
s
1
) in the wavelength interval
d (m), b
a.g.q..T
is the average absorption cross section of the gas in the interval
(cm
2
molec.
1
), and Y
q. p..T
is the average quantumyield of the photoprocess in the
interval (molec. photon
1
). Experimental absorption cross section and quantum
yield data vary with wavelength and temperature. References for such data are
given in Appendix Table B.4.
Photolysis of a molecule may produce one or more sets of products. Photolysis
of the nitrate radical, for example, produces two possible sets of products,
N
O
3
+h
NO
2
+
410 nm - - 670 nm
NO+O
2
590 nm - - 630 nm
(10.36)
The probability of each set of products is embodied in the quantum yield, dened
in Chapter 9. Whereas the absorption cross section of a gas is the same for each set
of products, the quantum yield of a gas differs for each set. Figures 10.3 (a) and
(b) show modeled photolysis-coefcient proles for several photoprocesses under
specied conditions.
10.5 SETS OF REACTIONS
Atmospheric chemical problems require the determination of gas concentrations
when many reactions occur at the same time. A difculty arises because a species
is usually produced and/or destroyed by several reactions. Consider the following
four reactions and corresponding rate expressions:
NO+O
3

NO
2
+O
2
Rate
1
= k
1
[NO][O
3
] (10.37)
+O
2
+M O
3
+M Rate
2
= k
2
[O][O
2
][M] (10.38)
NO
2
+h
NO+
Rate
3
= J [NO
2
] (10.39)
NO
2
+O
NO+O
2
Rate
4
= k
3
[NO
2
][O] (10.40)
Time derivatives of NO, NO
2
, O, and O
3
concentrations from the reactions are
d[NO]
dt
= P
c
L
c
= Rate
3
+Rate
4
Rate
1
= J [NO
2
] +k
3
[NO
2
][O] k
1
[NO][O
3
] (10.41)
351
10
5
0.0001 0.001 0.01 0.1 1
300
400
500
600
700
800
900
1000
Photolysis rate coefficient (s
1
)
P
r
e
s
s
u
r
e

(
h
P
a
)
b
NO
3
NO
2
a
O
3
a
NO
3
HONO
b
HCHO
a
HCHO
b
O
3
(a)
10
6
10
5
300
400
500
600
700
800
900
1000
Photolysis rate coefficient (s
1
)
P
r
e
s
s
u
r
e

(
h
P
a
)
HNO
3
CH
3
COCH
3
b
HO
2
NO
2
a
HO
2
NO
2
H
2
O
2
CH
3
COOH
CH
3
CHO
(b)
Figure 10.3 Modeled photolysis coefcient proles at 11:30 a.m. on August
27, 1987, at Temecula, California (33.49
N, 117.22
W). Cross-section and

quantum-yield data are referenced in Appendix Table B.4. The photoprocesses are
as follows:
a
O
3
+h O
2
+O
a
HCHO+h 2H+CO
b
O
3
+h O
2
+O(
1
D)
b
HCHO+h H
2
+CO
a
NO
3
+h NO
2
+ O
a
HO
2
NO
2
+h HO
2
+NO
2
b
NO
3
+h NO+O
2
b
HO
2
NO
2
+h OH+NO
3
NO
2
+h NO+O HONO+h OH+NO
CH
3
COCH
3
+h CH
3
+COCH
3
H
2
O
2
+h 2OH
HNO
3
+h OH+NO
2
CH
3
OOH+h CH
3
O+OH
The rate coefcient for HNO
3
decreases with increasing height near 400 hPa
because the absorption cross section of HNO
3
decreases with decreasing temper-
ature, and this factor becomes important at 400 hPa. The superscripts a and
b identify different sets of products for the same photodissociating molecule.
d[NO
2
]
dt
= P
c
L
c
= Rate
1
Rate
3
Rate
4
= k
1
[NO] [O
3
] J [NO
2
] k
3
[NO
2
] [O] (10.42)
d[O]
dt
= P
c
L
c
= Rate
3
Rate
2
Rate
4
= J [NO
2
] k
2
[O][O
2
][M] k
3
[NO
2
][O] (10.43)
d[O
3
]
dt
= P
c
L
c
= Rate
2
Rate
1
= k
2
[O][O
2
][M] k
1
[NO][O
3
] (10.44)
respectively. These equations are rst-order, rst-degree, homogeneous ordinary
differential equations (ODEs), as dened in section 6.1. In (10.38) and (10.40),
O
2
is not affected signicantly by the reactions, and no expression for the rate of
352
10.6 Stiff systems
change of O
2
is needed. Also, M is neither created nor destroyed, so no expression
for the rate of change of M is needed either. The O
2
and M concentrations are still
included in the rate expression of (10.38). In (10.41)(10.44), P
c
and L
c
are the
total rates of chemical production and loss, respectively. In the case of NO,
P
c
= J [NO
2
] +k
3
[NO
2
][O] L
c
= k
1
[NO][O
3
] (10.45)
Thousands of reactions occur in the atmosphere simultaneously. Because com-
puter resources are limited, an atmospheric models chemical mechanism needs to
be limited to the most important reactions. Whether a reaction is important can be
determined from an analysis of its rate. The rate of a reaction depends on the rate
coefcient and reactant concentrations. Rate coefcients vary with temperature,
pressure, and/or solar radiation. Reactant concentrations vary with time of day,
season, and location. A low rate coefcient and low reactant concentrations ren-
der a reaction unimportant. If the rate coefcient and reactant concentrations are
expected to be consistently small for a reaction during a simulation, the reaction
can be ignored.
10.6 STIFF SYSTEMS
Gas and aqueous chemical reaction sets are stiff. Stiff systems of reactions are more
difcult to solve numerically than nonstiff systems. A stiff system of reactions is
one in which the lifetimes (or time scales) of species taking part in the reactions
differ signicantly from one another. Species lifetimes are expressed as e-folding or
half-lifetimes. An e-folding lifetime is the time required for a species concentration
to decrease to 1e its original value. Ahalf-lifetime is the time required for a species
concentration to decrease to 12 its original value.
The overall chemical lifetime of a species is determined by calculating the lifetime
of the species against loss from individual reactions and applying
A
=
1
1
A1
+
1
A2
+
1
An
(10.46)
where
A
is the overall chemical lifetime of species A, and
A1
, . . . ,
An
are the
lifetimes of A due to loss from reactions 1, . . . , n, respectively. Equation (10.46)
applies to e-folding or half-lifetime calculations.
The e-folding lifetime of species A due to a unimolecular reaction of the form
A products is calculated fromd[A]dt = k
F
[A]. Integrating this equation from
concentration [A]
0
at t = 0 to [A] at t = h gives [A] = [A]
0
e
k
F
h
. The e-folding
lifetime is the time at which
[A]
[A]
0
=
1
e
= e
k
F
h
(10.47)
This occurs when
A1
= h =
1
k
F
(10.48)
353
In the case of a bi- or termolecular reaction, the e-folding lifetime is calculated
by assuming that each concentration, except for that of the species of interest,
equals the initial concentration of the species. Thus, the integration is reduced
to that of a unimolecular reaction. For instance, the rate of loss of A from the
bimolecular reaction A +B products can be described by the linearized ODE,
d[A]dt = k
S
[A][B]
0
, where k
S
is a second-order rate coefcient of the reaction,
and [B]
0
is the initial concentration of species B. This equation gives A an e-folding
lifetime of
A2
=
1
k
S
[B]
0
(10.49)
Similarly, the loss rate of A from the termolecular reaction A +B +C products
can be described by a linearized ODE that results in an e-folding lifetime for
A of
A3
=
1
k
T
[B]
0
[C]
0
(10.50)
where k
T
is the rate coefcient of the reaction, and [C]
0
is the initial concentration
of C.
The half-lifetime of a unimolecular reaction is determined in a manner similar
to its e-folding lifetime. The half-lifetime is the time at which
[A]
[A]
0
=
1
2
= e
k
F
h
(10.51)
This occurs when
(
1
/
2
)
A1
= h =
0.693
k
F
(10.52)
Analogous equations for bi- and termolecular reactions are, respectively,
(
1
/
2
)
A2
=
0.693
k
S
[B]
0
(
1
/
2
)
A3
=
0.693
k
T
[B]
0
[C]
0
(10.53)
Example 10.4
The e-folding lifetimes of atmospheric gases vary signicantly. The main loss
of methane is to the reaction CH
4
+OH CH
3
+H
2
O, where k
r
= 6.2
10
15
cm
3
molec.
1
s
1
at 298 K. When [OH] = 5.0 10
5
molec. cm
3
,
CH
4
=
1(k
r
[OH]) = 10.2 years.
For O(
1
D) +N
2
O+N
2
, k
r
= 2.6 10
11
cm
3
molec.
1
s
1
at 298 K.
When [N
2
] =1.9 10
19
molec. cm
3
, the e-folding lifetime of O(
1
D) is
O(
1
D)
=
1(k
r
[N
2
]) = 2 10
9
s.
Example 10.4 shows that the ratio of lifetimes between CH
4
and O(
1
D) is about
17 orders of magnitude. A system of equations that includes species with a wide
354
10.8 Problems
range of lifetimes is said to be stiff. Stiffness depends on reaction rates and species
concentrations. At low concentrations, some species lose their stiffness. For almost
all atmospheric cases, the range in species lifetimes is large enough for chemical
equations to be stiff.
10.7 SUMMARY
In this chapter, chemical species, structures, reactions, reaction rate coefcients,
and reaction rates were discussed. Elementary reactions are unimolecular, bimolec-
ular, or termolecular and give rise to rst-order, second-order, and third-order
rate coefcients, respectively. Photolysis reactions are unimolecular and give rise
to rst-order rate coefcients. Bimolecular reactions include thermal decomposi-
tion, isomerization, and basic collision reactions. Termolecular reactions include
combination and collision reactions. A combination reaction consists of a pair of
elementary bimolecular reactions. In the atmosphere, reactions occur simultane-
ously, and the lifetimes of species against chemical loss vary by orders of magni-
tude. Thus, ordinary differential equations describing atmospheric chemistry are
stiff.
10.8 PROBLEMS
10.1 WhenT=265 Kandp
d
=223 hPa, calculate the second-order rate coefcient
for
H+O
2
M
H
O
2
10.2 When T = 298 K and p
d
= 1013 hPa, calculate the rst-order rate coefcient
for
N
2
O
5
M

NO
2
+N
O
3
Repeat for T = 288 K. Discuss temperature effects on the reaction rate coef-
cient.
10.3 Estimate the e-folding lifetimes of CO, NO, O
3
, SO
2
, HNO
3
, ISOP (iso-
prene), and HO
2
against loss by OH if [OH] = 1.0 10
6
molec. cm
3
, T
= 288 K, and p
d
= 1010 hPa. The rate coefcients are listed in Appendix
Table B.4. Order the species from shortest to longest lifetimes. Which
species will most likely reach the stratosphere if only OH reaction is
considered?
10.4 Write rate expressions for the reactions
Cl +O
3
Cl
O+O
2
and Cl
O+

Cl +O
2
Write the time derivative of each species in the reactions, assuming the
reactions are solved together. What is the expression for the steady-state
concentration of ClO?
355
10.5 Given the following observed rate coefcients as a function of temperature,
ndthe activationenergy andcollisional prefactor of the associatedbimolec-
ular reaction. Showyour work. Inthe table, 1.303 (14) means 1.303 10
14
.
T (K) 278 288 298 308 318
k
r
(cm
3
molec.
1
s
1
) 1.303 (14) 1.547 (14) 1.814 (14) 2.106 (14) 2.422 (14)
10.6 Write a computer script to calculate the rst-order rate coefcient for
N
2
O
5
M
NO
2
+N
O
3
as a function of temperature and pressure. Drawgraphs of the rate coefcient
versus temperature when p
d
=1013 hPa and p
d
=800 hPa, respectively, and
versus pressure when T = 298 K and T = 275 K, respectively. Discuss the
results.
10.7 Write a computer script to read in reactions and rate-coefcient data from
a computer le. Use the script to calculate the rate coefcients for the
rst 15 reactions in Appendix Table B.4 when T = 288 K and p
d
=
980 hPa. Calculate reaction rates when [O
3
] = 2.45 10
12
, [O] = 1.0 10
3
,
[O(
1
D)] = 1.0 10
3
, [H] = 1.0 10
0
, [OH] = 1.0 10
6
, [H
2
] = 1.5 10
13
,
[HO
2
] = 1.0 10
8
, [N
2
O] = 7.6 10
12
, and [H
2
O] = 2.0 10
17
molec. cm
3
.
356
11
Urban, free-tropospheric, and stratospheric
chemistry
D
i fferent regions of the atmosphere are affected by different sets of chemical
reactions. The free troposphere is affected primarily by reactions among inor-
ganic, light organic, and some heavy organic gases. Urban regions are affected by
inorganic, light organic, and heavy organic gases. The stratosphere is affected pri-
marily by inorganic, light organic, and chlorinated/brominated gases. The impor-
tance of a reaction also varies between day and night and among seasons. In this
chapter, chemical reaction pathways are described for the free troposphere, urban
regions, and the stratosphere. Special attention is given to the marine sulfur cycle,
ozone production in urban air, and ozone destruction cycles in the global and
polar stratosphere. Heterogeneous reactions of gases on particle surfaces are also
described.
11.1 FREE-TROPOSPHERIC PHOTOCHEMISTRY
Photochemistry in the troposphere outside of urban regions is governed primar-
ily by reactions among inorganic and low-molecular-weight organic gases. High-
molecular-weight organic gases emitted anthropogenically, such as toluene and
xylene, break down chemically over hours to a few days, so they rarely penetrate
far fromurban regions. Air over the tropics and other vegetated regions, such as the
southeastern United States, is affected by reaction products of isoprene, a hemiter-
pene, and other terpenes emitted from biogenic sources. Reaction pathways for
these gases are discussed in Section 11.2. In the following subsections, inorganic
and low-molecular-weight organic reaction pathways for the free troposphere are
described.
11.1.1 Photostationary-state relationship
In many regions of the troposphere, the ozone (O
3
) mixing ratio is controlled by a
set of three tightly coupled chemical reactions involving itself, nitric oxide (NO),
and nitrogen dioxide (NO
2
). These reactions are
NO+O
3

NO
2
+O
2
(11.1)
NO
2
+h
NO+
- 420 nm (11.2)
+O
2
+M O
3
+M (11.3)
357
Urban, free-tropospheric, and stratospheric chemistry
In the free troposphere, the mixing ratios of O
3
(2060 ppbv) are much higher
than are those of NO (160 pptv) or NO
2
(570 pptv) (e.g., Singh et al.
1996), so (11.1) does not deplete ozone during day or night. At night in urban
regions, NO mixing ratios may exceed those of O
3
, and (11.1) can deplete local
ozone.
Assuming (11.1) and (11.2) have rate coefcients k
1
and J, respectively, the time
rate of change of the nitrogen dioxide concentration from these reactions is
d[NO
2
]
dt
= k
1
[NO][O
3
] J [NO
2
] (11.4)
If the time rate of change of NO
2
is small compared with those of the other terms
in (11.4), NO
2
is nearly in steady state and (11.4) simplies to the photostationary-
state relationship,
[O
3
] =
J [NO
2
]
k
1
[NO]
(11.5)
Equation (11.5) does not state that ozone in the free troposphere is affected by
only [NO] and [NO
2
]. Instead, it provides a relationship among [O
3
], [NO
2
], and
[NO]. If two of the three concentrations are known, the third can be found from
the equation.
Example 11.1
Find the photostationary-state mixing ratio of O
3
at midday when p
a
=
1013 hPa, T = 298 K, J 0.01 s
1
,
NO
= 5 pptv,
NO
2
= 10 pptv (typical
free-tropospheric mixing ratios), and the air is dry.
SOLUTION
At T = 298 K, k
1
1.810
14
cm
3
molec.
1
s
1
from Appendix Table B.4.
Since the conversion from mixing ratio to number concentration is the same
for each gas, [NO
2
][NO] =
NO
2
NO
. From (11.5), [O
3
] = 1.1 10
12
molec.
cm
3
. From (10.20), N
d
= N
a
= 2.4610
19
molec. cm
3
. Dividing [O
3
] by
N
d
gives
O
3
= 44.7 ppbv, which is a typical free tropospheric ozone mixing
ratio.
Two important reactions aside from (11.1)(11.3) that affect ozone are
O
3
+h O
2
+
O(
1
D) - 310 nm (11.6)
O
3
+h O
2
+
> 310 nm (11.7)

358
11.1 Free-tropospheric photochemistry
where O(
1
D) is excited atomic oxygen. In the free troposphere, the e-folding life-
times of ozone against destruction by these reactions are about 0.7 and 14 h,
respectively. The lifetime of ozone against destruction by NO in (11.1) was 126 h
under the conditions of Example 11.1. Thus, in the daytime free troposphere,
photolysis destroys ozone faster than does reaction with NO. When the ozone
concentration changes due to photolysis, the [NO
2
][NO] ratio in (11.5) changes
to adapt to the new ozone concentration.
The photostationary-state relationship in (11.5) is useful for free-tropospheric
analysis. In urban air, though, the relationship often breaks down because reactions
of NO with organic gas radicals provide an additional important source of NO
2
not included in (11.4). When organic-radical concentrations are large, as they are
during the morning in urban air, the photostationary-state relationship does not
hold. In the afternoon, though, organic gas concentrations in urban air decrease,
and the relationship holds better.
When excited atomic oxygen forms, as in (11.6), it rapidly produces O by
O(
1
D)
M
(11.8)
and O rapidly produces O
3
by (11.3). The reactions (11.3) and (11.6)(11.8) cycle
oxygen atoms quickly among O(
1
D), O, and O
3
. Losses of O from the cycle, such
as from conversion of NO to NO
2
by (11.1), are slower than are transfers of O
within the cycle.
11.1.2 Hydroxyl radical
The hydroxyl radical (OH) is an important chemical in the atmosphere because
it decomposes (scavenges) many gases. Its globally averaged tropospheric con-
centration is about 8 10
5
molec. cm
3
(Singh 1995). Its daytime concentra-
tion at any given location in the clean free troposphere ranges from 2 10
5
to
3 10
6
molec. cm
3
. When clean air is exposed to an urban plume, OH concen-
trations increase to 6 10
6
molec. cm
3
or more (Comes et al. 1997). In urban
air, OH concentrations range from 10
6
to 10
7
molec. cm
3
.
The primary free-tropospheric source of OH is
O(
1
D) +H
2
O 2
OH (11.9)
Sources of water vapor were discussed in Section 2.1.2.1. In the upper troposphere,
H
2
Ois scarce, limiting the ability of (11.9) to produce OH. Under such conditions,
other reactions, which are normally minor, may be important sources of OH. These
reactions are discussed in Section 11.1.10.
Minor sources of OH in the free troposphere are photolysis of gases, some of
which are produced by OH itself. For example, nitrous acid (HONO), produced
359
during the day by the reaction of NO with OH (Table 11.1), rapidly photolyzes
soon after by
HONO+h
OH+
NO - 400 nm (11.10)
HONO concentrations are high only during the early morning because the pho-
tolysis rate increases as sunlight becomes more intense during the day. During the
night, HONO is not produced from gas-phase reactions, since OH, required for
its production, is absent at night. HONO is produced from gas-particle reactions
and emitted from vehicles during day and night.
Nitric acid (HNO
3
), produced during the day by the reaction of NO
2
with OH
(Table 11.1), photolyzes slowly to reproduce OH by
HNO
3
+h
OH+
NO
2
- 335 nm (11.11)
The e-folding lifetime of nitric acid against destruction by photolysis is 1580 days,
depending on season and latitude. Since this lifetime is fairly long, HNO
3
is a
temporary reservoir of OH. Nitric acid is soluble, and much of it dissolves in cloud
or aerosol water and reacts on aerosol surfaces. HNO
3
mixing ratios are 5200 pptv
in the free troposphere.
Hydrogen peroxide (H
2
O
2
), produced by HO
2
self-reaction (Section 11.1.3),
photolyzes to OH by
H
2
O
2
+h 2
OH - 355 nm (11.12)
H
2
O
2
s e-folding lifetime against photolysis is 12 days. H
2
O
2
is soluble and is
often removed from the atmosphere by chemical reaction in clouds and dissolution
in precipitation. It is also lost by reaction with OH with an e-folding lifetime of
314 days in the free troposphere (Table 11.1).
Peroxynitric acid (HO
2
NO
2
), produced by the reaction of NO
2
with HO
2
(Section 11.1.3), photolyzes to produce OH by
HO
2
NO
2
+h
O
2
+
NO
2
- 330 nm
OH+N
O
3
- 330 nm
(11.13)
The e-folding lifetimes of both reactions are 25 days. HO
2
NO
2
is slightly soluble
and is removed by clouds and precipitation. It is also lost by reaction with OH
(Table 11.1) and thermal decomposition (Section 11.1.3).
After its formation, OHreacts with and decomposes many inorganic and organic
gases. The overall e-folding lifetime of OH against chemical destruction is about
0.11 s. The e-folding lifetimes of gases reacting with OH vary, as shown in
Table 11.1. The e-folding lifetimes of additional organic gases against destruction
by OH are given in Table 11.5 (Section 11.2).
360
Table 11.1 The e-folding lifetimes of several gases against destruction by OH in clean
free-tropospheric air when N
OH
is low and high
e-folding lifetime
N
OH
= 5 10
5
N
OH
= 2 10
6
Reaction molec. cm
3
molec. cm
3
Equation
OH+O
3
H
O
2
+O
2
346 d 86.5 d (11.14)
OH+H
2
H
2
O+
H 9.5 y 2.4 y (11.15)
OH+H
O
2
H
2
O+O
2
5.1 h 1.3 h (11.16)
OH+H
2
O
2
H
O
2
+H
2
O 13.6 d 3.4 d (11.17)
OH+
NO
M
HONO 2.4 d 14 h (11.18)
OH+
NO
2
M
HNO
3
1.9 d 11.4 h (11.19)
OH+HO
2
NO
2
H
2
O+
NO
2
+O
2
4.6 d 1.2 d (11.20)
OH+
SO
2
M
HS

O
3
26 d 6.5 d (11.21)
OH+CO
H+CO
2
111 d 28 d (11.22)
OH+CH
4
H
2
O+
CH
3
10.2 y 2.6 y (11.23)
OH+C
2
H
6
H
2
O+
C
2
H
5
93 d 23 d (11.24)
OH+C
3
H
8
H
2
O+
C
3
H
7
21 d 5.3 d (11.25)
OH+CH
3
OOH H
2
O+CH
3
O
2
6.4 d 1.6 d (11.26)
T=298 Kandp
a
=1013 hPa. HO
2
=hydroperoxy radical, H
2
O
2
=hydrogen peroxide, HONO=
nitrous acid, HNO
3
= nitric acid, NO
3
= nitrate radical, HO
2
NO
2
= peroxynitric acid, CH
3
=
methyl radical, HSO
3
= bisulte radical, C
2
H
6
= ethane, C
2
H
5
= ethoxy radical,
C
3
H
8
= propane, C
3
H
7
= propoxy radical, CH
3
OOH = methyl hydroperoxide, and
CH
3
O
2
methylperoxy radical.
11.1.3 Hydroperoxy radical
Like OH, the hydroperoxy radical (HO
2
) is a scavenger. Whereas OH is present
during daytime only, HO
2
is present during day and night. Concentrations of
HO
2
usually exceed those of OH. Chemical sources of HO
2
include OH + O
3
(Table 11.1), OH + H
2
O
2
(Table 11.1), photolysis of HO
2
NO
2
(11.13),
H+O
2
M
H
O
2
(11.27)
and the thermal decomposition of peroxynitric acid,
HO
2
NO
2
M
H
O
2
+
NO
2
(11.28)
This reaction occurs within seconds at high temperature near the surface, but
in the cold upper troposphere, it is slow. The H in (11.27) is produced primar-
ily from formaldehyde photolysis (Section 11.1.8.1) and the OH + CO reaction
(Table 11.1).
The rate of HO
2
loss depends on the NO mixing ratio. In the presence of
high NO (>10 pptv), HO
2
reacts mostly with NO and NO
2
. When NO is lower
361
(310 pptv), HO
2
reacts mostly with ozone. When NO is very low (-3 pptv),
HO
2
reacts mostly with itself (Finlayson-Pitts and Pitts 2000). The reactions
corresponding to these conditions are
H
O
2
+
NO
OH+
NO
2
H
O
2
+
NO
2
M
HO
2
NO
2
>10 pptv NO (11.29)

H
O
2
+O
3

OH+2O
2
310 pptv NO (11.30)
H
O
2
+H
O
2
H
2
O
2
+O
2
-3pptv NO (11.31)
11.1.4 Nighttime nitrogen chemistry
During the day in the free troposphere, NO and NO
2
are involved in the photosta-
tionary state cycle. Some losses from the cycle include the reactions OH+NO and
OH + NO
2
(Table 11.1). During the night, these loss processes shut down since
OH is absent at night. In addition, the photostationary relationship breaks down
because the photolysis reaction (11.2) shuts off, eliminating the source of O for
ozone production in (11.3) and the source of NO for ozone destruction in (11.1).
Because NO
2
photolysis shuts down, NO
2
becomes available at night to produce
the nitrate radical (NO
3
), dinitrogen pentoxide (N
2
O
5
), and aqueous nitric acid
(HNO
3
(aq)) by the sequence
NO
2
+ O
3
NO
3
+ O
2 (11.32)
M
NO
2
+ NO
3
N
2
O
5
(11.33)
N
2
O
5
+ H
2
O(aq) 2HNO
3
(aq)
(11.34)
Reaction (11.32) occurs during the day as well, but, during the day, it is less
important than NO + O
3
, and the NO
3
produced from the reaction is destroyed
almost immediately by sunlight.
Reaction (11.33) is a reversible reaction. The forward reaction is a three-body,
pressure-dependent reaction. The reverse reaction is a temperature-dependent ther-
mal decomposition reaction. At high temperature, such as during the day and in
the lower atmosphere, the reverse reaction occurs within seconds. At low temper-
ature, such as at night and at high altitudes, it occurs within hours to days or even
months.
Reaction (11.34) is a heterogeneous reaction, in that it involves a gas reacting
with a chemical on an aerosol particle or hydrometeor particle surface. In this case,
the chemical on the surface is liquid water. The reaction can occur on ice surfaces
as well, as described in Section 11.3.6. In the absence of liquid water or ice, the
corresponding homogeneous reaction (gas-phase only), N
2
O
5
+ H
2
O, can occur,
but the reaction is very slow.
362
After sunrise, NO
3
photolyzes almost immediately (with an e-folding lifetime of
seconds) by
N
O
3
+h
NO
2
+
410 nm - - 670 nm
NO+O
2
590 nm - - 630 nm
(11.35)
so NO
3
is not important during the day. Since N
2
O
5
is not produced during the day
and thermally decomposes within seconds by the reverse of (11.33) after sunrise,
it is also unimportant during the day. A slower daytime loss of N
2
O
5
is photolysis
(with an e-folding lifetime of hours),
N
2
O
5
+h
NO
2
+N
O
3
- 385 nm (11.36)
Another loss is mechanism heterogeneous reaction by (11.34).
11.1.5 Carbon monoxide production of ozone
The mixing ratio of ozone in the free troposphere, controlled primarily by the
photostationary-state relationship, is enhanced slightly by carbon monoxide (CO),
methane (CH
4
), and nonmethane organic gases. CO, with a typical tropospheric
mixing ratio of 100 ppbv, produces ozone by the sequence
CO+
OH CO
2
+
H (11.37)
H+O
2
M
H
O
2
(11.38)
NO+H
O
2

NO
2
+
OH (11.39)
NO
2
+h
NO+
- 420 nm (11.40)
+ O
2
+M O
3
+M (11.41)
Because the lifetime of CO against breakdown by (11.37) in the free troposphere
is 28110 days, the rate of ozone production by this sequence is slow. The mecha-
nism affects the photostationary-state relationship only slightly through Reaction
(11.39).
11.1.6 Methane production of ozone
Methane, with a mixing ratio of 1.8 ppmv, is the most abundant organic gas in
the Earths atmosphere. Table 11.1 indicates that its free-tropospheric e-folding
lifetime is about 10 years. This long lifetime has enabled it to mix uniformly up
to the tropopause. From this height upward, its mixing ratio gradually decreases.
Methanes only important loss is the reaction,
CH
4
+
OH
CH
3
+H
2
O (11.42)
363
This reaction, which produces the methyl radical (CH
3
), sets in motion the sequence
of reactions,
+ O
2
, M
C
H
H
H
Methyl
radical
C
H
H
H
O
O
Methylperoxy
radical
H
C O
H
H
Methoxy
radical
O
2
NO
2
+ NO
+HO
2
H C
H
O
H
O H
Methyl
hydroperoxide
C
O
H
Formaldehyde
+ O
2
HO
2
H
(11.43)
which leads to ozone production, but the incremental quantity of ozone produced
is small compared with the photostationary quantity of ozone.
The rst pathway of the methylperoxy radical (CH
3
O
2
) reaction produces NO
2
and formaldehyde (HCHO). Both produce ozone. NO
2
produces ozone by (11.2)
(11.3). Formaldehyde produces ozone as described in Section 11.1.8. The e-folding
lifetime of the methoxy radical (CH
3
O) intermediary against destruction by O
2
is
10
4
s; thus, its conversion to formaldehyde is almost instantaneous.
The second pathway of CH
3
O
2
reaction produces methyl hydroperoxide
(CH
3
OOH), which stores OH and HO
2
radicals. CH
3
OOH releases OH and
HCHO during photolysis and releases CH
3
O
2
during reaction with OH. The
e-folding lifetime of CH
3
OOH against photolysis is 1.52.5 days, and that against
OH reaction is 1.66.4 days at 298 K (Table 11.1). The methyl hydroperoxide
decomposition reactions are
H C
O
H
Formaldehyde
+
O
2
HO
2
H
C O
H OH
Methoxy
radical
C
H
O
H
O H
C
H
H
H
O
O
Methylperoxy
radical
Methyl
hydroperoxide
+h
< 360 nm
H
OH
H
2
O
H
+
(11.44)
364
11.1.7 Ethane and propane production of ozone
The primary nonmethane hydrocarbons in the free troposphere are ethane (C
2
H
6
),
propane (C
3
H
8
), ethene (C
2
H
4
), and propene (C
3
H
6
). Free-tropospheric mixing
ratios of these gases are 02.5 ppbv for ethane, 01.0 ppbv for propane, 0
1.0 ppbv for ethene, and 01.0 ppbv for propene (Singh et al. 1988; Bonsang
et al. 1991). The primary oxidant of ethane, propane, and other alkanes is OH.
Photolysis, reaction with O
3
, reaction with HO
2
, and reaction with NO
3
do not
affect alkane concentrations signicantly. In this subsection, ethane and propane
oxidation and production of ozone are discussed. Ethene and propene oxidation
pathways are described in Section 11.2.2.
The hydroxyl radical attacks ethane by
+
O
2
, M
Ethane
H
H
C C
H
H
H
H
H
C C
H
H
H
Ethyl radical
H
H
H
C C
H
O
H
H
O
Ethylperoxy radical
+
OH
H
2
O
(11.45)
The e-folding lifetime for the OH reaction is about 90 days in the free troposphere.
The ethylperoxy radical (C
2
H
5
O
2
) produced by the reaction takes one of two
courses,
H
H
C C
H
O
H
H
O
Ethylperoxy radical
H
H
C C
H
H
H
Ethoxy radical
O
NO
2
NO
Acetaldehyde
H
H
C C
H
O
H
+
O
2
HO
2
NO
2
M
H
H
C C
H
H
O
H
Ethylperoxynitric acid
O
N
O
O
+
+
+
(11.46)
The rst pathway leads to NO
2
, HO
2
, and acetaldehyde (CH
3
CHO). NO
2
pro-
duces ozone through (11.2)(11.3), HO
2
produces ozone through (11.39)(11.41),
and acetaldehyde produces ozone as described in Section 11.1.8. The ethoxy rad-
ical (C
2
H
5
O) produced in the rst pathway also reacts with NO to produce ethyl
nitrite (C
2
H
5
ONO) and with NO
2
to produce ethyl nitrate (C
2
H
5
ONO
2
), which
are storage reservoirs for nitrogen. Ethyl nitrate can also be produced directly by
365
reaction of the ethylperoxy radical with NO (not shown), but the e-folding lifetime
of C
2
H
5
O
2
against this reaction is relatively long (48192 days).
The second pathway leads to ethylperoxynitric acid (C
2
H
5
O
2
NO
2
), which is
a temporary storage reservoir for nitrogen since the reaction is reversible. The
e-folding lifetime of C
2
H
5
O
2
against loss by this reaction is 1.76.7 days.
The hydroxyl radical attacks propane by
+O
2
, M
H C
H
C
H
H
C
H
H
H
Propane n-Propyl radical
H
H C
H
C
H
H
C
H
H
O
H
n-Propylperoxy radical
O
OH
H
2
O
NO
2
n-Propoxy radical
O C
H
C
H
O
C
H
H
H
Acetone
+
O
2
HO
2
H C
H
C
H
H
C
H
H
H
H C
H
C
H
H
C
H
H
H
H
NO +
+
(11.47)
The e-folding lifetime of propane against loss by OH is about 20 days in the free
troposphere, more than four times shorter than the lifetime of ethane. The NO
2
from propane oxidation forms ozone by (11.2)(11.3). The acetone (CH
3
COCH
3
)
formed from(11.47) is a long-lived species whose fate is discussed in Section 11.1.9.
The propylperoxy radical (C
3
H
7
O
2
) competitively reacts with NO to form propyl
nitrate (C
3
H
7
ONO
2
), but the reaction is slow. The propoxy radical (C
3
H
7
O) reacts
with NOto formpropyl nitrite (C
3
H
7
ONO) and with NO
2
to formpropyl nitrate.
The free-tropospheric e-folding lifetimes of C
3
H
7
O against loss by these two reac-
tions are both 416 h.
11.1.8 Formaldehyde and acetaldehyde production of ozone
Formaldehyde is a carcinogen in high concentrations, an eye irritant, and an impor-
tant ozone precursor. It is a colorless gas with a strong odor at mixing ratios
>0.05 ppmv. Typical mixing ratios in urban air are -0.1 ppmv. It is the most
abundant aldehyde in the air and moderately soluble in water. It is produced chem-
ically from (11.43) and decomposes by photolysis, reaction with OH, reaction
with HO
2
, and reaction with NO
3
. Other sources include incomplete combustion
and emission from plywood, resins, adhesives, carpeting, particleboard, and ber-
board. Acetaldehyde is a precursor to ozone and peroxyacetyl nitrate (PAN). It is
produced by ethoxy-radical oxidation and destroyed by photolysis, and reaction
with OH, HO
2
, and NO
3
. Below, formaldehyde and acetaldehyde decomposition
is discussed.
366
11.1.8.1 Aldehyde photolysis
Formaldehyde and acetaldehyde photolyze during the day by
H C
O
H
Formaldehyde
+ h
+ H
Formyl
radical
CO + H
2
< 334 nm
< 370 nm
C
H O
(11.48)
C
H
H
H
C H
H
C C
H
Acetaldehyde Methyl radical Formyl radical
+
O
H
O
H
< 325 nm
+ h
(11.49)
respectively, producing H, CH
3
, and the formyl radical (HCO). Hreacts with O
2
to
form HO
2
by (11.27), which produces ozone by (11.39)(11.41). CH
3
is oxidized
in (11.43) eventually producing ozone. HCO produces CO and HO
2
by
Formyl
radical
C
O
H
CO
+
O
2
HO
2
(11.50)
CO and HO
2
form ozone by CO oxidation (Section 11.1.5).
11.1.8.2 Aldehyde reaction with the hydroxyl radical
Formaldehyde and acetaldehyde react with the hydroxyl radical to produce ozone
and peroxyacetyl nitrate, respectively. The formaldehyde process is
C
H C
Formaldehyde Formyl radical
O
H
O
H
+
OH
H
2
O
(11.51)
HCO produces CO in (11.50), which produces ozone as described in Section
11.1.5.
367
The acetaldehyde process is
O
H
H
H
H C C
Acetaldehyde Acetyl radical
O
H
H
H C C
Peroxyacetyl
radical
O
O
H
H
H C C
O
+
OH
H
2
O
+ O
2
, M
(11.52)
Peroxyacetyl
radical
C
O
O
H
H
H
O
Acetyloxy radical
O
O
H
H
H C C
Peroxyacetyl nitrate
C C
O
O
H
H
H
O
N
O
O
NO
2
NO +
+
C
NO
2
, M
+
(11.53)
The NO
2
formed from this process produces ozone. The process also produces
peroxyacetyl nitrate (PAN). PAN does not cause severe health effects, even in pol-
luted air, but it is an eye irritant and damages plants by discoloring their leaves.
Mixing ratios of PAN in clean air are typically 2100 pptv. Those in rural air
downwind of urban sites are up to 1 ppbv. Polluted air mixing ratios increase to
35 ppbv, with typical values of 10 to 20 ppbv. PAN mixing ratios peak during the
afternoon, the same time that ozone mixing ratios peak. PAN is not an important
constituent of air at night or in regions of heavy cloudiness. PAN was discovered
during laboratory experiments of photochemical smog formation (Stephens et al.
1956). Its only source is chemical reaction in the presence of sunlight. At 300 K
and at surface pressure, PANs e-folding lifetime against thermal decomposition is
about 25 minutes. At 280 K, its lifetime increases to 13 hours.
11.1.8.3 Aldehyde reaction with nitrate
At night, acetaldehyde reacts with NO
3
by the sequence
C C
O
H
H
H
H
Acetaldehyde Acetyl radical
C C
H
H
H
Peroxyacetyl
radical
C C
O
O
H
H
H
O
HNO
3
+O
2
, M
O
NO
3
+
(11.54)
368
Since the peroxyacetyl radical forms PAN, as shown in (11.53), Reaction (11.54)
is a nighttime source of PAN.
11.1.9 Acetone reactions
Acetone is a long-lived ketone produced from the OH oxidation of propane
(through 11.47), i-butane, or i-pentane (e.g., Chateld et al. 1987). The mixing
ratio of acetone in the free troposphere is 200700 pptv (Singh et al. 1995). Ace-
tone decomposes by reaction with OH and photolysis. The OH reaction produces
CH
3
COCH
2
and H
2
O. The e-folding lifetime of acetone against this reaction is
27 days when T=298 Kand [OH] =5 10
5
molec. cm
3
. The photolysis reaction
is
H C
H
C
H
O
C
H
H
H
Acetone
+ h
C
H
H
H
Methyl radical
+
Acetyl radical
C C
O
H
H
H
(11.55)
For the location and time given in Fig. 10.3(b), the e-folding lifetime of acetone
against photolysis is 1423 days, with the shorter lifetime corresponding to upper
tropospheric conditions and the longer lifetime corresponding to lower tropo-
spheric conditions. The acetyl radical from (11.55) forms PAN through (11.52)
(11.53). The methyl radical is oxidized in (11.43), producing HO
2
, HCHO, and
CH
3
OOH, among other products. HCHO produces HO
2
through (11.48) and
(11.50). CH
3
OOH decomposes, forming HO
2
and OH in the process, through
(11.44). Thus, acetone photolysis yields primarily PAN, OH, HO
2
, and interme-
diate products.
OH production from the O(
1
D)H
2
O reaction in the upper troposphere is less
signicant than in the lower troposphere because the concentration of H
2
O is
low in the upper troposphere. In the upper troposphere, acetone photolysis and
subsequent reaction may be an important source of OH (Singh et al. 1995).
11.1.10 Sulfur photochemistry
In the free troposphere, several naturally emitted gases contain sulfur (S). These
gases are important because they oxidize to sulfur dioxide, which is oxidized fur-
ther to sulfuric acid. Sulfuric acid condenses onto or forms new aerosol particles.
Aerosol particles affect radiation and serve as sites on which newcloud drops form,
and clouds affect radiation. Radiation affects temperatures, which affect pressures,
and winds. Thus, natural gases containing sulfur feed back to weather and climate.
Sulfur is emitted naturally in several forms. Anaerobic bacteria in marshes emit
dimethyl sulde (DMS) (CH
3
SCH
3
), dimethyl disulde (DMDS) (CH
3
SSCH
3
),
methanethiol (CH
3
SH), and hydrogen sulde (H
2
S). Phytoplankton in the oceans
369
emit DMS, DMDS, and other products. Volcanos emit carbonyl sulde (OCS),
carbon disulde (CS
2
), H
2
S, and sulfur dioxide (SO
2
).
Dimethyl sulde is produced by bacteria in some soils and plants, and it is the
most abundant sulfur-containing compound emitted from the oceans. DMS is pro-
duced from DMSP (dimethyl sulfonium propionate), which is emitted by many
phytoplankton (Bates et al. 1994). When phytoplankton feed, DMSP is exuded
and cleaved by enzymes to produce DMS and other products. Before DMS evap-
orates from the ocean surface, much of it chemically reacts or is consumed by
microorganisms. Over the oceans, DMS mixing ratios vary between -10 pptv and
1 ppbv. The latter value occurs over eutrophic waters (Berresheim et al. 1995).
Average near-surface DMS mixing ratios over oceans and land are about 100 and
20 pptv, respectively.
DMS is lost chemically in the atmosphere through hydroxyl radical abstraction
and addition. Abstraction is the process by which a radical, such as the hydroxyl
radical, removes an atom from a compound. Addition is the process by which a
radical bonds to a compound. Above 285 K, hydroxyl radical abstraction is the
dominant DMSOH reaction pathway. Below 285 K, addition is more important
(Hynes et al. 1986). Yin et al. (1990) and Tyndall and Ravishankara (1991) present
detailed oxidation pathways of DMS.
Hydroxyl radical abstraction of DMS results in the formation of the methanethi-
olate radical (CH
3
S) and formaldehyde by the sequence
+
O
2
H
H
S
H
C C
H
H
H H C
H
S
H
C
H
H
H
H
S
H
C C
H
H
O
O
Dimethyl sulfide (DMS) DMS radical DMS peroxy radical
H
2
O
H
H
S
H
C C
H
H
O H C
H
S
H
+ H C
O
H
DMS oxy radical Methanethiolate
radical
Formaldehyde
M
NO
2
+ NO
OH
+
(11.56)
The methanethiolate radical reacts with O
2
by
+
O
2
, M
H C
H
S
H
H C
H
S
H
O
O
Methanethiolate
radical
Excited methanethiolate
peroxy radical
*
Methanethiolate oxy
radical
H C
H
S
H
O
NO
2
+ NO
(11.57)
370
where the intermediate product is short-lived. The methanethiolate oxy radical
(CH
3
SO) decays to sulfur monoxide (SO) or forms the methanethiolate peroxy
radical (CH
3
SO
2
) by
H C
H
H
Methyl
radical
H C
H
S
H
O
+
H C
H
S
H
O
O
S O
Sulfur monoxide
Methanethiolate
peroxy radical
Methanethiolate oxy
radical
M
+
O
3
O
2
(11.58)
Sulfur monoxide forms sulfur dioxide by
S
O O
O
Sulfur
monoxide
S O
Sulfur
dioxide
+ O
2
(11.59)
The e-folding lifetime of sulfur monoxide against destruction by O
2
is about
0.0005 s at 298 K and 1 atm pressure. CH
3
SO
2
from (11.58) forms sulfur dioxide
by
S
O O
H
H
H
C + H
H
S C
H
O
O
Methyl radical Sulfur dioxide Methanethiolate
peroxy radical
M
(11.60)
CH
3
SO
2
also reacts with O
3
, NO
2
, and HO
2
to produce CH
3
SO
3
, which either
breaks down to CH
3
+ SO
3
or abstracts a hydrogen atom from an organic gas to
form CH
3
S(O)
2
OH (methanesulfonic acid, MSA). In sum, OH abstraction of
DMS results in the formation of sulfur dioxide, methanesulfonic acid, and other
products.
371
The DMS-addition pathway initiates when OH bonds to the sulfur atom in
DMS. The reaction sequence is
H
H
S
H
C C
H
H
H
+OH
Dimethyl sulfide (DMS)
H
H
S
H
C
H
H
H
OH
H C
H
H
H
H
S
H
O C
H
O
S C
O
C
H
H
H
H
H
H
+
Dimethyl sulfone (DMSO
2
)
Methanesulfenic
acid
Methyl
radical
DMS-OH adduct
M
+OH, 2O
2
2HO
2
C
(11.61)
The dimethyl sulfone from this sequence does not react further. The methane-
sulfenic acid (CH
3
SOH) produces CH
3
SO by
Methanethiolate oxy
radical
H
H
S
H
O
H
Methanesulfenic
acid
H
H
S
H
O
+OH
H
2
O
C C
(11.62)
CH
3
SO reacts via (11.58)(11.60) to produce sulfur dioxide.
Dimethyl disulde (DMDS) oxidation proceeds almost exclusively by OH addi-
tion. The sequence is
H C
H
S
H
S C
H
H
H H C
H
S
H
O
H
H C
H
S
H
+
Dimethyl disulfide (DMDS) Methanesulfenic
acid
Methanethiolate
radical
+
OH
(11.63)
CH
3
SOH and CH
3
S react in (11.62) and (11.57), respectively, to form SO
2
, MSA,
and other products. DMDS also photolyzes by
H C
H
S
H
S C
H
H
H
H C
H
S
H
2
Dimethyl disulfide (DMDS)
Methanethiolate
radical
+ h
(11.64)
372
Hydrogen sulde (H
2
S), which has the odor of rotten eggs, is emitted from
anaerobic soils, plants, paper manufacturing sources, and volcanos. It is also pro-
duced in the deep ocean but does not evaporate from surface ocean water before
it is oxidized. Over the remote ocean, H
2
S mixing ratios range from 5 to 15 pptv.
Over shallow coastal waters, where more evaporation occurs, its mixing ratios
reach 100300 pptv. Over land, its mixing ratios are 5150 pptv. Downwind of
industrial sources, they increase to 1100 ppbv (Berresheim et al. 1995). In the air,
hydrogen sulde reacts with OH by
S
H H
S
H
Hydrogen

sulfide

+OH
H
2
O Hydrogen
sulfide radical
(11.65)
to form the hydrogen sulde radical (HS), which subsequently reacts with O
3
,
NO
2
, and O
2
. The e-folding lifetimes of HS against loss by these three gases are
0.27 s, 14 s, and 0.49 s, respectively when T = 298 K, p
d
= 1013 hPa, [O
3
] =
40 ppbv, and [NO
2
] = 50 pptv. The HSO
2
reaction is
+
O
2
S
H
OH Hydrogen
sulfide radical
Sulfur
monoxide
S O
(11.66)
Sulfur monoxide produces SO
2
by (11.59). The HSO
3
and HSNO
2
reactions
produce HSO, which reacts again with O
3
and NO
2
to produce HSO
2
, which
reacts further with O
2
to form SO
2
and HO
2
.
Methanethiol (methyl sulde, CH
3
SH), emitted by bacteria, reacts with OH to
produce the methanethiolate radical by
H
H
S
H
H
H
S
H
H
Methanethiol Methanethiolate
radical
+OH
H
2
O
C
C
(11.67)
CH
3
S initiates SO
2
production in (11.57).
Carbonyl sulde (OCS) and carbon disulde (CS
2
) are emitted from volcanos.
Both react with OH, but the OCS-OH reaction is slow, and photolysis of OCS
occurs only in the stratosphere. OCS mixing ratios have increased over time and,
at 500 pptv, are the highest among background sulfur-containing compounds.
Mixing ratios of OCS are relatively uniform between the surface and tropopause,
373
at which point photolysis reduces them. The OH and photolysis reactions of OCS
are, respectively,
+ OH O C S
S
H
Hydrogen sulfide
radical
+ CO
2
Carbonyl
sulfide
(11.68)
Carbonyl
sulfide
O C S
Carbon
monoxide
Atomic
sulfur
+ h
< 260 nm CO S
+
(11.69)
Atomic sulfur from (11.69) forms SO by
O
+ O
2
Sulfur
monoxide
S O
Atomic
sulfur
S
(11.70)
SO forms SO
2
by (11.59).
Carbon disulde is broken down by reaction with OH and photolysis. Because
the CS
2
OH and CS
2
photolysis reactions are faster than are the OCSOH and
OCS photolysis reactions, respectively, CS
2
mixing ratios of 2200 pptv are lower
than are OCS mixing ratios in the background atmosphere. The CS
2
OH and CS
2
photolysis reactions are
+ OH S C S
S
H
O S
+
Carbon
disulfide
Hydrogen sulfide
radical
Carbonyl
sulfide
C
(11.71)
S C
C
S +
+
+
h
S
S
Carbon
disulfide
Carbon
monosulfide
Atomic
sulfur
< 340 nm
(11.72)
respectively. Carbon monosulde (CS) produces OCS by
+ O
2
C
Carbon
monosulfide
+ O O C S
Carbonyl
sulfide
S
+
(11.73)
Once sulfur dioxide has been produced chemically in or emitted into the atmo-
sphere, it nds its way into aerosol particles and clouds by one of two mechanisms.
The rst is dissolution of SO
2
into water-containing aerosol particles and cloud
drops followed by its aqueous oxidation to sulfuric acid (H
2
SO
4
(aq)). This mech-
anism is discussed in Chapter 19. The second mechanism is gas-phase oxidation of
374
11.2 Urban photochemistry
SO
2
to sulfuric acid gas (H
2
SO
4
), which readily condenses onto aerosol particles to
form aqueous sulfuric acid. The gas phase conversion of sulfur dioxide to sulfuric
acid requires three steps. The rst is the conversion of SO
2
to bisulte (HSO
3
).
Bisulte quickly reacts with oxygen to form sulfur trioxide (SO
3
). Sulfur trioxide
then reacts with water to form sulfuric acid gas. The reaction sequence is
S
O O
O
S
HO O
O
S
O O
+ OH, M + H
2
O
Sulfur
dioxide
Bisulfite
Sulfur
trioxide
Sulfuric
acid
S
O
OH
O OH
+O
2
HO
2
(11.74)
With both mechanisms, the aqueous sulfuric acid formed dissociates into ions, as
described in Chapter 17.
In sum, free-tropospheric chemistry is governed by inorganic and light organic
reactions. Important inorganic constituents in the free troposphere are NO, NO
2
,
O
3
, OH, HO
2
, CO, and SO
2
. Organic species of interest include CH
4
, C
2
H
6
,
C
3
H
8
, C
2
H
4
, C
3
H
6
, HCHO, CH
3
CHO, and CH
3
COCH
3
. In the free troposphere,
the relationship among NO, NO
2
, and O
3
can be quantied reasonably with the
photostationary-state relationship. Some gases in the free troposphere originate
from natural sources while others originate from combustion and other emission
sources. Urban photochemistry is discussed next.
11.2 URBAN PHOTOCHEMISTRY
Two general categories of urban air pollution have been observed. The rst is
London-type smog and the second is called photochemical smog.
London-type smog results from the burning of coal and other raw materials in
the presence of a fog or a strong temperature inversion. The pollution consists of a
high concentration of directly emitted particles containing metals, ash, and acids or,
when fog is present, fog drops laden with these constituents. Several deadly London-
type smog events occurred in London in the nineteenth and twentieth centuries,
including one in 1952 in which 4000 deaths above average were recorded. Deadly
events in Glasgow and Edinburgh, Scotland (1909, 1000 deaths), Meuse Valley,
Belgium (December, 1930, 63 deaths), and Donora, Pennsylvania (October, 1948,
20 deaths) have been attributed to London-type smog.
Photochemical smog results from the emission of organic gases and oxides of
nitrogen in the presence of sunlight. The pollution consists of a soup of gases and
aerosol particles. Some of the particles are directly emitted, whereas others are
produced by gas-to-particle conversion. Today, many cities exhibit photochemi-
cal smog, including, among others, Mexico City, Santiago, Los Angeles, Tehran,
375
Calcutta, Beijing, Tokyo, Johannesburg, and Athens. All these cities experience
some degree of London-type together with photochemical smog.
11.2.1 Basic characteristics of photochemical smog
Photochemical smog differs from background air in two ways. First, mixing ratios
of nitrogen oxides and organic gases are higher in polluted air than in background
air, causing ozone levels to be higher in urban air than in the background. Second,
photochemical smog contains higher concentrations of high molecular weight
organic gases, particularly aromatic gases, than does background air. Because such
gases break down quickly in urban air, most are unable to survive transport to the
background troposphere.
Photochemical smog involves reactions among nitrogen oxides (NO
x
= NO +
NO
2
) and reactive organic gases (ROGs, total organic gases minus methane) in the
presence of sunlight. The most recognized gas-phase by-product of smog reactions
is ozone because ozone has harmful health effects and is an indicator of the presence
of other pollutants.
On a typical morning, NOand ROGs are emitted by automobiles, power plants,
and other combustion sources. Emitted pollutants are primary pollutants. ROGs
are oxidized to organic peroxy radicals, denoted by RO
2
, which react with NO
to form NO
2
. Pre-existing ozone also converts NO to NO
2
. Sunlight then breaks
down NO
2
to NO and O. Finally, O reacts with molecular oxygen to form ozone.
The basic reaction sequence is thus
NO+R
O
2

NO
2
+R
O (11.75)
NO+O
3

NO
2
+O
2
(11.76)
NO
2
+h
NO+
- 420 nm (11.77)
+ O
2
+M O
3
+M (11.78)
Pollutants, such as ozone, that formchemically or physically in the air are secondary
pollutants. NO
2
, which is emitted and forms chemically, is both a primary and
secondary pollutant.
Because RO
2
competes with O
3
to convert NOto NO
2
in urban air, and because
the photostationary state relationship is based on the assumption that only O
3
converts NOto NO
2
, the photostationary relationship is usually not valid in urban
air. In the afternoon, the relationship holds better than it does in the morning
because RO
2
mixing ratios are lower in the afternoon than in the morning.
Figure 11.1 shows ozone mixing ratios resulting from different initial mixtures
of NO
x
and ROGs. This type of plot is an ozone isopleth. The isopleth shows that,
at lowNO
x
, ozone is relatively insensitive to ROGs levels. At high NO
x
, an increase
in ROGs increases ozone. Also, at low ROGs, increases in NO
x
above 0.05 ppmv
decrease ozone. At high ROGs, increases in NO
x
always increase ozone.
376
0 0.5 1 1.5 2
0
0.05
0.1
0.15
0.2
0.25
ROG (ppmC)
N
O
x

(
p
p
m
v
)
0. 4
0
.
3
2
0
.
2
4
0
.
1
6
0
.
0
8

=

O
3
,

p
p
m
v
Figure 11.1 Peak ozone mixing ratios result-
ing from different initial mixing ratios of NO
x
and ROGs. The ROG:NO
x
ratio along the line
through zero is 8:1. Adapted fromFinlayson-Pitts
and Pitts (2000).
An isopleth is useful for regulatory control of ozone. If ROG mixing ratios are
high (e.g., 2 ppmC) and NO
x
mixing ratios are moderate (e.g., 0.06 ppmv), the plot
indicates that the most effective way to reduce ozone is to reduce NO
x
. Reducing
ROGs under these conditions has little effect on ozone. If ROG mixing ratios
are low (e.g., 0.7 ppmC), and NO
x
mixing ratios are high (e.g., 0.2 ppmv), the
most effective way to reduce ozone is to reduce ROGs. Reducing NO
x
under these
conditions increases ozone. In many polluted urban areas, the ROG:NO
x
ratio is
lower than 8:1, indicating that limiting ROG emission should be the most effective
method of controlling ozone. Because ozone mixing ratios depend not only on
chemistry but also on meteorology, washout, dry deposition, and gas-to-particle
conversion, such a conclusion is not always clearcut.
11.2.2 Meteorological factors affecting smog
Meteorological factors also affect the development of air pollution; Los Angeles is
a textbook example. The Los Angeles basin is bordered on its southwestern side by
the Pacic Ocean and on all other sides by mountain ranges. During the day, a sea
breeze blows inland. The sea breeze is at its strongest in the afternoon, when the
temperature difference between land and ocean is the greatest. At night, a reverse
land breeze (from land to sea) occurs, but it is often weak. Figure 11.2(a) shows
the variation of sea- and land-breeze wind speeds at Hawthorne, which is near the
coast in the Los Angeles basin, during a three day period in 1987.
The sea breeze is instrumental in advecting primary pollutants, emitted mainly
on the west side of the Los Angeles basin, toward the east side, where they arrive
as secondary pollutants. During transport, primary pollutants, such as NO, are
converted to secondary pollutants, such as O
3
. Whereas NO mixing ratios peak
on the west side of Los Angeles, as shown in Fig. 11.2(b), O
3
mixing ratios peak
377
0
1
2
3
4
5
6
7
0 6 12 18 24 30 36 42 48 54 60 66 72
W
i
n
d

s
p
e
e
d

(
m

s
1
)
Hour of day
Day 1
Day 2
Day 3
(a)
0
0.1
0.2
0.3
0 6 12 18 24
V
o
l
u
m
e

m
i
x
i
n
g

r
a
t
i
o

(
p
p
m
v
)
Hour of day
O
3
NO
2
NO
(b)
0
0.1
0.2
0.3
0 6 12 18 24
V
o
l
u
m
e

m
i
x
i
n
g

r
a
t
i
o

(
p
p
m
v
)
Hour of day
O
3
NO
2
NO
(c)
Figure 11.2 (a) Wind speeds at Hawthorne from August 26 to 28, 1987.
The other panels show the evolution of the NO, NO
2
, and O
3
mixing
ratios at (b) central Los Angeles and (c) San Bernardino on August 28.
Central Los Angeles is closer to the coast than is San Bernardino. As the
sea breeze picks up during the day, primary pollutants, such as NO, are
transported fromthe western side of the Los Angeles basin (e.g., central Los
Angeles) toward the eastern side (e.g., San Bernardino). As the pollution
travels, organic peroxy radicals convert NO to NO
2
, which forms ozone,
a secondary pollutant.
on the east side, as shown in Fig. 11.2(c). The west side of the basin is a source
region and the east side is a receptor region of photochemical smog.
Other factors that exacerbate air pollution in the Los Angeles basin are its
location relative to the Pacic high-pressure systemand its exposure to sunlight. The
Pacic high suppresses vertical air movement, inhibiting clouds other than stratus
from forming. The subsidence inversion caused by the high prevents pollution
from rising easily over the mountains surrounding the basin. Also, because the
basin is further south (about 34
N latitude) than most United States cities, Los

Angeles receives more daily radiation than do most cities, enhancing its rate of
photochemical smog formation relative to other cities.
11.2.3 Emission of photochemical smog precursors
Gases emitted in urban air include nitrogen oxides, reactive organic gases, carbon
monoxide, and sulfur oxides (SO
x
= SO
2
+ SO
3
). Of these, NO
x
and ROGs are
378
Table 11.2 Gas-phase emission for August 27, 1987 in a 400 150 km
region of the Los Angeles basin
Substance Emission (tons day
1
) Percentage of total
Carbon monoxide (CO) 9796 69.3
Nitric oxide (NO) 754
Nitrogen dioxide (NO
2
) 129
Nitrous acid (HONO) 6.5
Total NO
x
+ HONO 889.5 6.3
Sulfur dioxide (SO
2
) 109
Sulfur trioxide (SO
3
) 4.5
Total SO
x
113.5 0.8
Alkanes 1399
Alkenes 313
Aldehydes 108
Ketones 29
Alcohols 33
Aromatics 500
Hemiterpenes 47
Total ROGs 2429 17.2
Methane (CH
4
) 904 6.4
Total Emission 14 132 100
Source: Allen and Wagner (1992).
Table 11.3 Organic gases emitted in the greatest quantity in the Los Angeles basin
1. Methane 10. Propylene 19. Acetone 28. Methylcyclohexane
2. Toluene 11. Chloroethylene 20. n-Pentadecane 29. Nonane
3. Pentane 12. Acetylene 21. Cyclohexane 30. Methylalcohol
4. Butane 13. Hexane 22. Methylethylketone 31. 1-Hexane
5. Ethane 14. Propane 23. Acetaldehyde 32. Methylcyclopentane
6. Ethene 15. Benzene 24. Trimethylbenzene 33. Methylpentane
7. Octane 16. Methylchloroform 25. Ethylbenzene 34. Dimethylhexane
8. Xylene 17. Pentene 26. Methylvinylketone 35. Cyclopentene
9. Heptane 18. n-Butylacetate 27. Naphtha
Source: Pilinis and Seinfeld (1988).
the main precursors of photochemical smog. Table 11.2 shows emission rates of
several pollutants in the Los Angeles basin on a summer day in 1987. CO was the
most abundantly emitted gas. About 85 percent of NO
x
was emitted as NO, and
almost all SO
x
was emitted as SO
2
. Of the ROGs, toluene, pentane, butane, ethane,
ethene, octane, and xylene were emitted in the greatest abundance (Table 11.3).
The abundance of a gas does not necessarily translate into proportional smog
production. A combination of abundance and reactivity is essential for an ROG to
be an important smog producer.
379
Table 11.4 Percentage emission of several gases
by source category in Los Angeles in 1987
Source category CO NO
x
SO
x
ROG
Stationary 2 24 38 50
Mobile 98 76 62 50
Total 100 100 100 100
Source: Chang et al. (1991).
Table 11.4 shows the percentage emission of several gases by source cate-
gory in Los Angeles. Emissions originate from point, area, and mobile sources.
A point source is an individual pollutant source, such as a smokestack, xed
in space. A mobile source is a moving individual pollutant source, such as the
exhaust of a motor vehicle or an airplane. An area source is an area, such as
a city block, an agricultural eld, or an industrial facility, over which many
xed sources aside from smokestacks exist. Together, point and area sources
are stationary sources. Table 11.4 shows that CO, the most abundantly emit-
ted gas in the basin, originated almost entirely (98 percent) from mobile sources.
Oxides of nitrogen were emitted mostly (76 percent) by mobile sources. The
thermal combustion reaction in automobiles that produces nitric oxide at a high
temperature is
N
N+O == O+heat 2
N == O (11.79)
Table 11.4 also shows that stationary and mobile sources each accounted for
50 percent of ROGs emitted in the basin. Mobile sources accounted for 62 percent
of SO
x
emission. The mass of SO
x
emission was one-eighth that of NO
x
emission.
Sulfur emission in Los Angeles is lowrelative to that in many other cities worldwide
because Los Angeles has relatively few coal-red power plants, which are heavy
emitters of SO
x
.
11.2.4 Breakdown of ROGs
Once organic gases are emitted, they are broken down chemically into free rad-
icals. Six major processes break down hydrocarbons and other ROGs photol-
ysis and reaction with OH, HO
2
, O, NO
3
, and O
3
. Reaction of organics with
OH and O occurs only during the day, because OH and O require photolysis for
their production and are short-lived. NO
3
is present only at night because it pho-
tolyzes quickly during the day. O
3
and HO
2
may be present during both day and
night.
OH is produced in urban air by some of the same reactions that produce it in
the free troposphere. An early morning source of OH in urban air is photolysis
of HONO. Since HONO may be emitted by automobiles, it is more abundant in
380
urban air than in the free troposphere. Midmorning sources of OHin urban air are
aldehyde photolysis and oxidation. The major afternoon source of OH is ozone
photolysis. In sum, the three major reaction mechanisms that produce the hydroxyl
radical in urban air are
Early morning source
HONO+h
OH+
NO - 400 nm (11.80)
Midmorning source
HCHO+h
H+H
CO - 334 nm (11.81)
H+O
2
M
H
O
2
(11.82)
H
CO+O
2
H
O
2
+CO (11.83)
NO+H
O
2

NO
2
+
OH (11.84)
Afternoon source
O
3
+h O
2
+
O(
1
D) - 310 nm (11.85)
O(
1
D) +H
2
O 2
OH (11.86)
ROGs emitted in urban air include alkanes, alkenes, alkynes, aldehydes, ketones,
alcohols, aromatics, and hemiterpenes. Table 11.5 shows lifetimes of these ROGs
against breakdown by six processes. The table shows that photolysis breaks down
aldehydes and ketones, OH breaks down all eight groups during the day, HO
2
breaks down aldehydes during the day and night, O breaks down alkenes and
terpenes during the day, NO
3
breaks down alkanes, alkenes, aldehydes, aromatics,
and terpenes during the night, and O
3
breaks down alkenes and terpenes during
the day and night.
The breakdown of ROGs produces radicals that lead to ozone formation.
Table 11.6 shows the most important ROGs in Los Angeles during the summer of
1987 in terms of a combination of abundance and reactive ability to form ozone.
The table shows that m- and p-xylene, both aromatic hydrocarbons, were the
most important gases in terms of generating ozone. Although alkanes are emitted
in greater abundance than are other organics, they are less reactive in producing
ozone than are aromatics, alkenes, or aldehydes.
In the following subsections, photochemical smog processes involving the chem-
ical breakdown of organic gases to produce ozone are discussed.
11.2.5 Ozone production from alkanes
Table 11.6 shows that i-pentane and butane are the most effective alkanes with
respect to the combination of concentration and reactivity in producing ozone in
381
Table 11.5 Estimated lifetimes of reactive organic gases representing alkanes, alkenes,
alkynes, aldehydes, ketones, aromatics, and terpenes against photolysis and oxidation in
urban and free-tropospheric air
Lifetime in polluted urban air at sea level
[OH] [HO
2
] [O] [NO
3
] [O
3
]
5 10
6
2 10
9
8 10
4
1 10
10
5 10
12
ROG species Photolysis molec. cm
3
molec. cm
3
molec. cm
3
molec. cm
3
molec. cm
3
n-Butane 22 h 1000 y 18 y 29 d 650 y
trans-2-Butene 52 m 4 y 6.3 d 4 m 17 m
Acetylene 3.0 d 2.5 y 200 d
Toluene 9.0 h 6 y 33 d 200 d
Isoprene 34 m 4 d 5 m 4.6 h
Formaldehyde 7 h 6.0 h 1.8 h 2.5 y 2.0 d 3200 y
Acetone 23 d 9.6 d
Lifetime in free-tropospheric air at sea level
[OH] [HO
2
] [O] [NO
3
] [O
3
]
5 10
5
3 10
8
3 10
3
5 10
8
1 10
12
ROG Species Photolysis molec. cm
3
molec. cm
3
molec. cm
3
molec. cm
3
molec. cm
3
n-Butane 9.2 d 6700 y 480 y 1.6 y 3250 y
trans-2-Butene 8.7 h 27 y 168 d 1.3 h 1.4 h
Acetylene 30 d 67 y 2.7 y
Toluene 3.8 d 160 y 1.8 y 2.7 y
Isoprene 5.7 h 106 d 1.7 h 23 d
Formaldehyde 7 h 2.5 d 11.7 h 67 y 40 d 16,000 y
Acetone 23 d 96 d
Estimated lifetimes for some species in urban air were recalculated from Finlayson-Pitts and
Pitts (2000). Lifetimes of other species were obtained from rate-coefcient data. Photolysis rate
coefcients were obtained from Figs. 10.3(a) and (b). Gas concentrations are typical, but not
necessarily average values for each region. Units: m, minutes; h, hours; d, days; y, years; , no
data or insignicant loss.
Table 11.6 Ranking of the most abundant species in terms of reactivity during the summer
Southern California Air Quality Study in 1987
1. m- and p-Xylene 8. o-Xylene 15. m-Ethyltoluene 22. p-Ethyltoluene
2. Ethene 9. Butane 16. Pentanal 23. C
4
Olen
3. Acetaldehyde 10. Methylcyclopentane 17. Propane 24. 3-Methylpentane
4. Toluene 11. 2-Methylpentane 18. Propanal 25. o-Ethyltoluene
5. Formaldehyde 12. Pentane 19. i-Butane
6. i-Pentane 13. 1,2,4-Trimethylbenzene 20. C
6
Carbonyl
7. Propene 14. Benzene 21. Ethylbenzene
Source: Lurmann et al. (1992). The ranking was determined by multiplying the weight fraction of
each organic present in the atmosphere by a species-specic reactivity scaling factor developed by
Carter (1991).
382
Los Angeles air. As in the free troposphere, the main pathway of alkane decompo-
sition in urban air is OH attack. Photolysis and reaction with O
3
, HO
2
, and NO
3
have little effect on alkane concentrations. Of all alkanes, methane is the least
reactive and the least important with respect to urban air pollution. Methane is
more important with respect to free-tropospheric and stratospheric chemistry. The
oxidation pathway of methane was given in Section 11.1.6, and those of ethane
and propane were given in Section 11.1.7.
11.2.6 Ozone production from alkenes
Table 11.6 shows that alkenes, such as ethene and propene, are important ozone
precursors in photochemical smog. Table 11.5 indicates that alkenes react most
rapidly with OH, O
3
, and NO
3
. In the following subsections, these reaction path-
ways are discussed.
11.2.6.1 Alkene reaction with the hydroxyl radical
When ethene reacts with the hydroxyl radical, the radical substitutes into ethenes
double bond to produce an ethanyl radical in an OH addition process. The ethanyl
radical then reacts to produce NO
2
(g). The sequence is
C C
H
H
H
H
Ethene
C C
H
H
H
H
Ethanyl radical
OH
C C
H
H
H
H
Ethanolperoxy
radical
OH
O
O
C C
H
H
H
H
Ethanoloxy
radical
OH
O
+O
2
, M
NO
2
+NO
OH, M +
(11.87)
NO
2
produces ozone by (11.2)(11.3). The ethanoloxy radical (HOCH
2
CH
2
O),
a by-product of ethene oxidation, produces formaldehyde and glycol aldehyde
(HOCH
2
CHO) by
Formaldehyde
C O
H
H
C C
H
H
H
H
Ethanoloxy
radical
OH
O
C C
H
H
O
H
OH
Glycol aldehyde
72% 2
28%
O
2
HO
2
+
(11.88)
383
Formaldehyde decomposition produces ozone, as discussed in Section 11.1.8.1.
Like other aldehydes, glycol aldehyde is decomposed by photolysis and reaction
with OH.
11.2.6.2 Alkene reaction with ozone
When ethene or propene reacts with ozone, the ozone substitutes into ethenes
double bond to form an unstable ethene or propene molozonide. The molozonide
quickly decomposes to products that are also unstable. The reaction of ethene with
ozone is
C C
H
H
H
H
+O
3
H
2
C
2
O
O O
Ethene Ethene molozonide
+
37%
Formaldehyde Criegee biradical
+
63%
Formaldehyde Excited criegee
biradical
C O
H
H
C O
H
H
C O
H
H
O
C O
H
H
O
*
CH
(11.89)
Formaldehyde produces ozone as described in Section 11.1.8. The criegee biradical
forms NO
2
by
Formaldehyde Criegee biradical
C O
H
H
C O
H
H
O
NO
2
+ NO
(11.90)
The excited criegee biradical isomerizes, and its product, excited formic acid, ther-
mally decomposes by
Excited criegee
biradical
C O
H
H
O
*
C O
H
O
*
60% CO + H
2
O
21% CO
2
+ H
2
19%
+ O
2
CO + OH + HO
2 Excited formic
acid
H
(11.91)
where the fractions are valid at room temperature (Atkinson et al. 1997). In sum,
ozone attack on ethene produces HCHO, HO
2
, CO, and NO
2
. These gases not
only reform the original ozone lost, but also produce new ozone.
384
Ozone oxidizes propene by the sequence
C CH
2
H
H
3
C
Propene
+ O
3
CH CH
2
O
O O
Propene molozonide
7.5%
42.5%
18.5%
31.5%
Formaldehyde Methyl criegee
biradical
+
Formaldehyde
Excited methyl
criegee biradical
C O
H
H
C O
H
3
C
H
O
*
H
3
C
C O
H
3
C
H
C O
H
3
C
H
Acetaldehyde
Acetaldehyde
Criegee biradical
Excited criegee
biradical
C O
H
H
O
C
H
H
O
*
+
C
H
H
C O
H
3
C
H
O
+
+ O
O
(11.92)
(Atkinson et al. 1997). The methyl criegee biradical reacts with NO by
C O C O
Acetaldehyde Methyl criegee
biradical
H
3
C
H
H
3
C
H
O
NO
2
+ NO
(11.93)
The excited methyl criegee biradical isomerizes to excited acetic acid, which ther-
mally decomposes via
Excited methyl criegee
biradical
C O
H
3
C
H
O
*
C O
H
3
C
O
*
16% CH
4
+ CO
2

64% CH
3
+ CO + OH
20% CH
3
O + HO
2
+ CO
Excited acetic
acid
H
(11.94)
where the fractions are valid at room temperature (Atkinson et al. 1997). Thus,
propene oxidation by ozone produces OH, HO
2
NO
2
, HCHO, and CH
3
CHO, all
of which react further to reform ozone.
385
11.2.6.3 Alkene reaction with nitrate
Table 11.5 shows that the reaction of NO
3
with an alkene is rapid. Such breakdown,
which occurs only at night, leads to a morning buildup of organic peroxy radicals.
The reaction sequences with respect to ethene and propene oxidation by NO
3
are
C C
H
H
H
H
+ NO
3
C CH
2
H
H
O N
O
O
+ O
2
C CH
2
H
H
O N
O
O
O
Ethylperoxy nitrate
radical
Ethyl nitrate radical
O
Ethene
+
+
C CH
2
H
H
O N
O
O
O
Ethoxy nitrate radical
NO
2
NO
+
+
(11.95)
+ NO
3
+ O
2
C C
H
H
3
C
H
H
C CH
2
H
H
3
C
O N
O
O CH
2
H
H
3
C
O N
O
O
O
Propylperoxy nitrate
radical
Propyl nitrate radical
C
O
Propene
+
+
NO
2
+ NO
C CH
2
H
H
3
C
O N
O
O
O
Propoxy nitrate radical
(11.96)
respectively. NO
2
produced from these reactions during the night leads to ozone
formation in the morning.
11.2.7 Ozone production from aromatics
Toluene (C
6
H
5
CH
3
) originates from gasoline combustion, biomass burning,
petroleum rening, detergent production, paint, and building materials. After
methane, it is the second most abundantly emitted organic gas in Los Angeles
air and the fourth most important gas in terms of abundance and chemical reactiv-
ity (Table 11.6). Mixing ratios of toluene in polluted air range from 1 to 30 ppbv.
386
Table 11.5 shows that toluene is decomposed almost exclusively by OH, which
breaks down toluene by abstraction and addition. The respective pathways are
CH
3
H
OH
O
O
CH
3
H
OH
CH
3
OH
H
2
C O O
CH
2
Toluene
o-Cresol
Benzylperoxy
radical
Toluene-hydroxyl-
radical adduct
Benzyl
radical
8%
92%
o-Hydroxytoluene
CH
3
+
O
2
+
O
2
+ HO
2
+ OH
H
2
O
+
OH
(11.97)
The benzylperoxy radical from the abstraction pathway reacts with NO by
+ NO
CH H
2
C
Benzaldehyde
Benzoxy
radical
Benzyl nitrate
H
2
C O O
Benzylperoxy
radical
H
2
C O N
O
O
O O
NO
2
+ NO
+ O
2
HO
2
+
(11.98)
387
Benzaldehyde, like formaldehyde and acetaldehyde, produces ozone. The toluene-
hydroxyl-radical adduct from the toluene addition pathway forms NO
2
by
CH
3
H
OH
Toluene-hydroxyl
radical adduct
CH
3
H
OH
O
O
O

NO2
+
NO
(11.99)
Cresol, from toluene addition, reacts with OH by
CH
3
OH
o-Cresol
CH
3
O
Methylphenylperoxy
radical
O
+ OH, 2O
2
2HO
2
CH
3
O
CH
3
OH
m-Nitrocresol
+ NO
2
N
O
Methylphenoxy
radical
+ OH
H
2
O
O
+
(11.100)
The methylphenylperoxy radical converts NO to NO
2
. Nitrocresol readily con-
denses onto particles, acting as a sink for NO
x
.
11.2.8 Ozone production from terpenes
The free troposphere and urban areas are affected by biogenic emission of iso-
prene and other terpenes. Biogenic emissions are emissions produced from biolog-
ical sources, such as plants, trees, algae, bacteria, and animals. Strictly speaking,
terpenes are hydrocarbons that have the formula C
10
H
16
. Loosely speaking, they
are a class of compounds that include hemiterpenes (C
5
H
8
), such as isoprene;
monoterpenes (C
10
H
16
), such as -pinene, -pinene, and d-limonene; sesquiter-
penes (C
15
H
24
); and diterpenes (C
20
H
32
). Isoprene is emitted by sycamore, oak,
aspen, spruce, willow, balsam, and poplar trees; -pinene is emitted by pines, rs,
cypress, spruce, and hemlock trees; -pinene is emitted by loblolly pine, spruce,
redwood, and California black sage trees; and d-limonene is emitted by loblolly
pine, eucalyptus, and California black sage trees and by lemon fruit.
388
Table 11.5 shows that OH, O
3
, and NO
3
decompose isoprene. Although iso-
prene does not have a long chemical lifetime, some of its by-products last longer.
11.2.8.1 Terpene reaction with the hydroxyl radical
The reaction pathways of isoprene with OH produce at least six peroxy radicals.
The pathways are
Isoprene
H
C C
H
2
C
CH
2
CH
3
CH C
C
H
2
CH
2
CH
3
HO
O
O
CH C
C
H
2
CH
2
CH
3
O
HO
O
H
C C
C
H
2
C
H
2
CH
3
HO
O
O
H
C C
H
2
C
C
H
2
CH
3
O
O
HO
H
C C
H
2
C
C
H
2
CH
3
OH
O
O
H
C C
C
H
2
C
H
2
CH
3
O
O
OH
(1)
16.4%
(2)
12.3%
(3)
12.3%
(4)
23.6%

(5)
21.2%
(6)
14.1%
Isoprene peroxy radicals
+
OH, O
2
(11.101)
(Paulson and Seinfeld 1992). The e-folding lifetime of isoprene against reaction
with OHis about 30 minutes when [OH] =5.0 10
6
molec. cm
3
. All six radicals
convert NO to NO
2
. The second and fth radicals also create methacrolein and
methylvinylketone by
CH C
C
H
2
CH
2
CH
3
O
HO
O
H
C C
O
CH
2
CH
3
Isoprene peroxy radical Methacrolein
+
Formaldehyde
C O
H
H
NO
2
+
NO +
O
2
HO
2
(11.102)
NO
2
+ O
2
HO
2
C O
H
H
H
C C
H
2
C
O
CH
3
Isoprene peroxy radical Methylvinylketone
+
Formaldehyde
H
C C
H
2
C C
H
2
CH
3
OH
O
O
NO
+
(11.103)
389
respectively. The NO
2
from these reactions produces ozone. Methacrolein and
methylvinylketone react with OH and O
3
to form additional products that convert
NO to NO
2
, resulting in more ozone, as shown in Appendix Table B.4.
11.2.8.2 Terpene reaction with ozone
When O
3
reacts with isoprene, it attacks isoprene at either end of either double
bond:
Isoprene
+ O
3
H
C C
H
2
C
CH
2
CH
3
H
C C
O
CH
2
CH
3
Methylvinylketone
H
C C
H
2
C
O
CH
3
H
C C
O
CH
2
CH
3
O
H
C C
H
2
C O
CH
3
O
+
+
+
+
Criegee biradical
Formaldehyde
Methacrolein
Ozonide product
Ozonide product Formaldehyde
Criegee biradical
C O
H
H
O
C O
H
H
O
C O
H
H
C O
H
H
(11.104)
The criegee biradical reacts further by (11.90). Formaldehyde reacts as shown in
(11.48) and (11.51). Both reactions produce by-products that form ozone.
11.2.9 Ozone production from alcohols
Two alternative motor-vehicle fuels are methanol and ethanol. Methanol oxidation
produces formaldehyde and ozone, and ethanol oxidation produces acetaldehyde,
a precursor to PAN. The reaction of methanol with OH is
H
H
O
H
Methanol
H
H O
H
H
H C
Formaldehyde
O
H
H
C O
H
H
Methoxy radical
85%
15%
+ OH
H
2
O
+ O
2
HO
2
C
C
(11.105)
390
Methanol has an e-folding lifetime against reaction with OH of 71 days when
[OH] = 5.0 10
6
molec. cm
3
; thus, the reaction is not rapid. The organic prod-
uct of the rst reaction is formaldehyde, and that of the second reaction is the
methoxy radical, which produces formaldehyde by (11.43). Formaldehyde is an
ozone precursor.
Ethanol oxidation by OH produces the branched reactions
H
H
O C C
H
Ethanol
H
C C
Acetaldehyde
O
H
H
C O
H
C
Ethoxy radical
5%
90%
H
H
H
H
H
H
H
H
H C C
H
H
H
5%
O C C
H
H
O
H
H
H
H
+OH
H
2
O
+O
2
HO
2
(11.106)
Ethanol lost from the most-probable (middle) reaction has an e-folding lifetime of
about 19 hours when [OH] = 5.0 10
6
molec. cm
3
. Acetaldehyde, formed from
the middle reaction, produces PAN and ozone. Cities in Brazil have experienced
high PAN mixing ratios since the introduction of their alcohol-fuel program.
11.2.10 Condensed mechanisms for organic chemistry
The number of chemical reactions involving organic gases in urban air is large.
Explicit chemical mechanisms with thousands of organic reactions have been devel-
oped (e.g., Madronich and Calvert 1989; Jenkin et al. 2003; Saunders et al. 2003).
Although such mechanisms can now be solved in a three-dimensional atmospheric
model for a period of a few days (e.g., Liang and Jacobson 2000), the compu-
tational demand for long-term and most practical simulations requires that the
number of species and reactions be reduced.
Three methods of reducing the number of organic reactions in a model are the
carbon-bond lumping method (e.g., Whitten et al. 1980; Gery et al. 1989), the
surrogate-species method (e.g., Atkinson et al. 1982; Lurmann et al. 1987; Grifn
et al. 2002), and the lumped-species method (e.g., Stockwell 1986; Carter 1990,
2000).
With the carbon-bond lumping method, individual organic gases are segregated
into one or more bond groups that have similar chemical reactivity. For example,
a butane molecule, which has four carbons connected by single bonds, is divided
into four single carbon atoms, each represented by the parafn (PAR) bond group.
391
Table 11.7 Carbon-bond representations of several organic gases
Chemical name
Carbon bond group
Chemical structure
Ethane
0.4 PAR + 1.6 UNR
H C
H
C
H
H
H
H
Ethene
1 ETH
C C
H
H
H
H
Formaldehyde
1 FORM
H C
O
H
Toluene
1 TOL
CH
3
Benzene
1 PAR + 5 UNR

2,2,4-Trimethylpentane
8 PAR
H
3
C C
H
CH
3
C
H
2
C
CH
3
CH
3
CH
3
Propene
1 PAR + 1 OLE
C CH
2
H
H
3
C
Propionaldehyde
1 PAR + 1 ALD2
H
C CH
2
O CH
3
m-Xylene
1 XYL
CH
3
CH
3
Cyclopentene
1 PAR + 2 ALD2
Cyclopentane
5 PAR
H
2
C
H
2
C
C
H
2
CH
2
H
2
C
Ethyne
1 PAR + 1 UNR
C C H H
Benzaldehyde
1 ALD2 + 5 UNR
CH O
1,2,3-Trimethylbenzene
1 PAR + 1 XYL
CH
3
CH
3
CH
3
Cyclohexene
2 PAR + 2 ALD2
n-Butane
4 PAR
Trans 2-butene
2 ALD2
H C
H
C
H
C
H
C
H
H
H
H
C
O
H
Acetaldehyde
1 ALD2
H C
H
H
Ethylbenzene
1 PAR + 1 TOL
CH
2
H
3
C
Methylethylketone
3 PAR + 1 KET
H
2

C C
H
3
C
O
CH
3
H C
H
C
H
H
C
H
H
H
C
H
H
H
Cyclopentane, which has ve single-bonded carbons, is broken into ve PAR bond
groups. All PAR bonds are assumed to have the same chemical reactivity, regard-
less of whether they originated from butane or cyclopentane. A terminal carbon
atom pair with a double bond between the two atoms is represented by an olen
(OLE). Nonterminal carbon-atom pairs with a double bond attached to one of
the carbons and terminal two-carbon carbonyl groups [CC(=O)H] are lumped
as ALD2. Single-carbon ketone groups (C=O) are lumped as KET, seven-carbon
392
11.3 Stratospheric photochemistry
aromatics are simulated as toluene (TOL), eight-carbon aromatics are simulated as
m-xylene (XYL), and terpenes are grouped as isoprene (ISOP). Methane (CH
4
),
ethene (ETH), formaldehyde (FORM), methanol (MEOH), ethanol (ETOH),
acetone (AONE), and several other species are not lumped. Nonreactive carbon
atoms in certain organic gases are labeled unreactive (UNR). In some cases, carbon
atoms with a double or triple bond that have similar reactivity to carbon atoms
with a single bond are labeled as PAR. Table 11.7 gives the carbon-bond repre-
sentation of a few organic gases. Several of the reactions in Appendix Table B.4
include reactions among carbon-bond species.
With the surrogate-species method, all species of similar reactivity are grouped
together. Propane and pentane are assumed to have the same reactivity as n-butane,
and all three species are grouped as one surrogate species. With the lumped-species
method, species of similar reactivity are lumped together, just as with the surrogate
species method. The difference is that with the surrogate-species method the reac-
tion rate coefcient for each surrogate species is set equal to that of a particular
gas. The reaction rate coefcient of a lumped species is determined before a model
simulation by taking a mole-fraction-weighted average of the reaction rates of each
species in the lumped group.
11.2.11 Summary of urban photochemistry
Photochemical smog production is governed by emission of oxides of nitrogen
and reactive organic gases. Emitted gases, called primary pollutants, react in the
presence of sunlight to produce secondary pollutants, such as ozone and peroxy-
acetyl nitrate. The radicals that break down emitted reactive organic gases are OH,
HO
2
, O
3
, NO
3
, and O. Photolysis also breaks down certain organics. Because
reactive organic gas radicals compete with O
3
to produce NO
2
from NO, the
photostationary-state relationship does not usually hold in urban air. Because gas-
phase organic chemistry involves reactions among thousands of species, condensed
reaction mechanisms have been developed to simplify the simulation of organic
chemistry in numerical models.
11.3 STRATOSPHERIC PHOTOCHEMISTRY
Whereas ozone molecules in urban air are harmful to humans, animals, plants,
trees, and structures, the same ozone molecules in the stratosphere are bene-
cial in that they shield the Earth from harmful ultraviolet radiation. Figure 11.3
shows a typical variation of ozone mixing ratio (
O
3
), ozone number concentration
(N
O
3
=
O
3
N
d
), and dry air number concentration (N
d
) with altitude. The ozone
number concentration (molecules of ozone per cubic centimeter of air) in the strato-
sphere generally peaks at 25 to 32 km altitude. The ozone mixing ratio peaks at a
higher altitude than does the ozone number concentration. The peak ozone number
concentration in the stratosphere is close to that in polluted urban air. The peak
ozone mixing ratio in the stratosphere (near 10 ppmv) is much higher than is that
in polluted urban air (0.20.35 ppmv) or free-tropospheric air (0.020.04 ppmv).
393
0 2 4 6 8 10 12 14
0
10
20
30
40
A
l
t
i
t
u
d
e

(
k
m
)
O
3
(ppmv)
O
3
(molecules cm
3
10
12
)
Air (molecules cm
3
5 10
19
)
Figure 11.3 Example vertical varia-
tion in ozone mixing ratio, ozone
number concentration, and air num-
ber concentration with altitude. The
ozone mixing ratio at the surface is
0.20 ppmv, the level of a Stage 1 smog
alert in the United States.
Approximately 90 percent of all ozone molecules in the atmosphere reside in the
stratosphere because the stratosphere is approximately 3240 km thick. Pollutants
in urban regions generally mix to depths of only 0.31.5 km.
Another measure of ozone is its column abundance, which is the sum of all
ozone molecules above a square centimeter of surface between the ground and the
top of the atmosphere. When this number is divided by 2.7 10
16
, the result is
the column abundance in Dobson units (DUs). Thus, 1 DU is equivalent to 2.7
10
16
molecules of ozone per square centimeter of surface (or 0.001 atm cm). In
2000, the globally-averaged column abundance of ozone from 90
S to 90
N was
293.4 DU. This column abundance contains the same number of molecules as a col-
umn of air 2.93-mmhigh at 1 atmof pressure and 273 K (near-surface conditions),
as illustrated in Fig. 11.4.
11.3.1 Ozone formation from oxygen
The altitude of the stratospheric ozone concentration peak in Fig. 11.3 occurs
where the concentration of oxygen and the quantity of ultraviolet radiation are
sufciently high to maximize ozone formation (Section 2.2.2.2). Above the ozone
peak, the concentration of oxygen is too low to produce peak ozone. Below the
peak, the quantity of radiation is too lowto produce peak ozone. In this subsection,
the oxygen cycle involved in producing the stratospheric ozone peak is revisited.
Amajor difference between ozone formation in the stratosphere and troposphere
is that ozone formation in the stratosphere is driven by the photolysis of molec-
ular oxygen whereas that in the troposphere is driven by photolysis of nitrogen
dioxide. Nitrogen dioxide photolysis is unimportant in the stratosphere because
NO
2
concentrations there are too low to produce much ozone. Oxygen photolysis
394
Figure 11.4 Example of globally-averaged column abun-
dance of ozone. The number of ozone molecules per unit
area of surface in a 293-DU column of ozone is equiva-
lent to the number of air molecules in a 2.93-mm high
column near the surface. (The gure is not to scale.)
is unimportant in the troposphere because ultraviolet radiation intensity required
to break down O
2
there is too weak.
Ozone formation in the stratosphere is initiated by the oxygen photolysis
reactions,
O
2
+h
O(
1
D) +
- 175 nm (11.107)
O
2
+h
+

O
175 - - 245 nm (11.108)

The rst reaction is important only at the top of the stratosphere, since wavelengths
shorter than 0.175 mdo not penetrate deeper. The second reaction occurs down to
the lower stratosphere, but not to the troposphere since wavelengths below245 nm
do not reach the troposphere. Since the oxygen concentration decreases and radi-
ation intensity increases with increasing height, an altitude exists where the pro-
duction of atomic oxygen is maximized. This is the altitude where ozone peaks in
Fig. 11.3.
As in the troposphere, ozone in the stratosphere forms from
+ O
2
+MO
3
+M (11.109)
Most of the atomic oxygen for this reaction comes directly from molecular oxygen
photolysis. Some of it comes from quenching of excited atomic oxygen to the
ground state by
O(
1
D)
M
(11.110)
395
In the upper stratosphere, most excited atomic oxygen comes from (11.107). In
the rest of the stratosphere (and in the troposphere), most of it comes from the rst
ozone photolysis reaction,
O
3
+h O
2
+
(
1
D) - 310 nm (11.111)
O
3
+h O
2
+
> 310 nm (11.112)

These photolysis reactions help to limit ozone formation primarily in the upper
stratosphere, where radiation intensity is high, but also in the lower stratosphere.
11.3.2 Effect of nitrogen on the natural ozone layer
Oxides of nitrogen naturally destroy ozone in the upper stratosphere, helping to
shave the vertical ozone prole. In the troposphere, the major sources of NO are
surface emission and lightning. The major source of NOin the stratosphere is trans-
port from the troposphere and the breakdown of nitrous oxide (N
2
O) (laughing
gas) by
N
2
O+
O(
1
D)
64% 2
NO
36% N
2
+O
2
(11.113)
The O(
1
D) for this reaction comes mostly from ozone and oxygen photolysis.
N
2
O is a colorless gas emitted by bacteria in fertilizers, sewage, and the oceans
and during biomass burning and automobile combustion. It affects not only strato-
spheric ozone, but also climate because it is a greenhouse gas. Its mixing ratio up to
1520 km is constant (0.31 ppmv) before decreasing. About 10 percent of nitrous
oxide loss in the stratosphere is due to (11.113) and the rest is due to photolysis by
N
2
O+h N
2
+
O(
1
D) - 240 nm (11.114)
which gives N
2
O an e-folding lifetime of about 1.3 yr at 25 km.
NO naturally reduces ozone in the upper stratosphere by
NO+O
3

NO
2
+O
2
(11.115)
NO
2
+
NO+O
2
(11.116)
+O
3
2O
2
(net process) (11.117)
This sequence destroys one molecule of ozone, but no molecules of NO or NO
2
. It
is called the NO
x
catalytic ozone destruction cycle because ozone is lost but NO
x
(NO + NO
2
) is recycled. The number of times the cycle is executed before NO
x
is removed from the cycle by reaction with another gas is the chain length. In the
upper stratosphere, the chain length of this cycle is about 10
5
(Lary 1997). Thus,
10
5
molecules of O
3
are destroyed before one NO
x
molecule is removed from the
cycle. In the lower stratosphere, the chain length decreases to near 10.
396
In the absence of chlorine, the major reactions removing NO
x
from the cycle
are
NO
2
+
OH
M
HNO
3
(11.118)
HO
2
+NO
2
HO
2
NO
2
(11.119)
The products of these reactions, nitric acid and peroxynitric acid, photolyze back to
the reactants that formed them by (11.11) and (11.13), respectively, but the rates
are slow. Peroxynitric acid also decomposes thermally by (11.28), but thermal
decomposition is slow in the stratosphere because temperatures are low there.
The natural NO
x
catalytic cycle erodes the ozone layer above ozones peak
altitude shown in Fig. 11.3. Although the NO
x
catalytic cycle is largely natural,
an unnatural source of stratospheric NO
x
and, therefore, ozone destruction, is
stratospheric aircraft emission of NO and NO
2
. In the 1970s and 1980s, scientists
were concerned that the introduction of a eet of supersonic transport (SST) jets
into the stratosphere would enhance NO
x
sufciently to damage the ozone layer.
However, the plan to introduce a eet of stratospheric jets never materialized.
11.3.3 Effect of hydrogen on the natural ozone layer
Hydrogen-containing compounds, particularly OH and HO
2
, are responsible for
shaving the ozone prole in the lower stratosphere. The hydroxyl radical is pro-
duced in the stratosphere by several reactions,
O(
1
D) +
H
2
O 2
OH
CH
4

CH
3
+
OH
H
2

H+
OH
(11.120)
OH participates in the HO
x
catalytic ozone destruction cycle, where HO
x
=
OH + HO
2
. The most effective HO
x
cycle, which has a chain length in the lower
stratosphere of 1 to 40 (Lary 1997), is
OH+O
3
H
O
2
+O
2
(11.121)
H
O
2
+O
3

OH+2O
2
(11.122)
2O
3
3O
2
net process (11.123)
HO
x
can be removed temporarily from catalytic cycles by (11.118), (11.119),
and
H
O
2
+
OH H
2
O+O
2
(11.124)
This mechanism is particularly efcient because it removes two HO
x
molecules at
a time.
397
10
5
0
5
1980 1985 1990 1995 2000
P
e
r
c
e
n
t

d
i
f
f
e
r
e
n
c
e

i
n

g
l
o
b
a
l

o
z
o
n
e
f
r
o
m

1
9
7
9

m
o
n
t
h
l
y

a
v
e
r
a
g
e
Year
Mount Pinatubo
(June, 1991)
El Chichon
(April, 1982)
Figure 11.5 Percentage change in the monthly
averaged global (90
S to 90
N) ozone column
abundance between a given month since 1979 and
the same month in 1979.
11.3.4 Effect of carbon on the natural ozone layer
Carbon monoxide and methane produce ozone in the stratosphere by Reactions
(11.37)(11.44). The contributions of CO and CH
4
to ozone production in the
stratosphere, though, are small. A by-product of methane oxidation in the strato-
sphere is water vapor, produced by
CH
4
+
OH
CH
3
+H
2
O (11.125)
Because the mixing ratio of water vapor in the stratosphere is low and transport
of water vapor from the troposphere to stratosphere is slow, this reaction is a
relatively important source of water in the stratosphere.
11.3.5 Chlorine and bromine photochemistry
Between 1979 and 2000, the global stratospheric ozone column abundance
decreased by approximately 3.5 percent (from 304.0 to 293.4 DU), as shown in
Fig. 11.5. Unusual decreases in global ozone occurred following the El Chich on
(Mexico) volcanic eruption in April 1982, and the Mount Pinatubo (Philippines)
eruption in June 1991. These eruptions injected particles into the stratosphere. On
the surfaces of these particles, chemical reactions involving chlorine took place that
contributed to ozone loss. Over time, however, the concentration of these particles
decreased, and the global ozone layer partially recovered. Because volcanic parti-
cles were responsible for only temporary ozone losses, the net loss of ozone over
the globe from 1979 to 2000 was still about 3.5 percent.
Between 1950 and 1980, no measurements from three ground-based stations
in the Antarctic showed ozone levels less than 220 DU, a threshold for den-
ing Antarctic ozone depletion. Every Southern Hemisphere spring (September
November) since 1980, measurements of stratospheric ozone have shown a deple-
tion. Farman et al. (1985) rst reported depletions of more than 30 percent
398
100
150
200
250
300
350
400
450
500
90 60 30 30 60 90
O
z
o
n
e

(
D
o
b
s
o
n

u
n
i
t
s
)
Latitude (degrees)
1999
1979
2000
0
Figure 11.6 Variation with latitude of
October monthly and zonally averaged col-
umn abundances of ozone in 1979, 1999,
and 2000. Data were obtained from the
satellite-based Total Ozone Mapping Spec-
trometer (TOMS) and made available by
NASA Goddard Space Flight Center, Green-
belt, Maryland. No data were available
from December 1994 to July 1996.
relative to pre-1980 measurements. Since then, measurements over the South Pole
have indicated depletions of up to 70 percent of the column ozone for a period of
a week in early October. The largest average depletion for the month of Septem-
ber from 60 to 90
S since 1979 was 32.8 percent and occurred in 2000. The

largest depletion for the month of October from 60 to 90
S since 1979 was

38.3 percent and occurred in 1998 (Fig. 11.6). Most ozone depletion has occurred
between altitudes of 12 and 20 km. The large reduction of stratospheric ozone
over the Antarctic in the Southern Hemisphere spring each year is the Antarctic
ozone hole. The areal extent of the ozone hole is nowgreater than the size of North
America.
11.3.5.1 Chlorouorocarbons and related compounds
Molina and Rowland (1974) recognized that anthropogenic chlorine compounds
could destroy stratospheric ozone. Since then, scientists have strengthened the links
among global ozone reduction, Antarctic ozone depletion, and the presence of
chlorine and other halogenated compounds in the stratosphere.
The compounds that play the most important role in reducing strato-
spheric ozone are chlorouorocarbons (CFCs). Important CFCs are identied in
Table 11.8. CFCs are gases formed synthetically by replacing all hydrogen atoms in
methane (CH
4
) or ethane (C
2
H
6
) with chlorine and/or uorine atoms. For exam-
ple, CFC-12 (CF
2
Cl
2
) is formed by replacing the four hydrogen atoms in methane
with two chlorine and two uorine atoms.
CFC-12 was invented in 1928 by Thomas Midgley and his assistants on the same
day that a representative of General Motors Frigidaire division asked Midgley to
nd a nontoxic, nonammable substitute for an existing refrigerant, ammonia, a
399
Table 11.8 Mixing ratios and lifetimes of selected chlorocarbons, bromocarbons,
and uorocarbons
Tropospheric Estimated overall
Chemical mixing atmospheric
formula Trade name Chemical name ratio (pptv) lifetime (yrs)
Chlorocarbons and chlorine compounds
Chlorouorocarbons (CFCs)
CFCl
3
CFC-11 Trichlorouoromethane 270 45
CF
2
Cl
2
CFC-12 Dichlorodiuoromethane 550 100
CFCl
2
CF
2
Cl CFC-113 1-Fluorodichloro- 70 85
2-diuorochloroethane
CF
2
ClCF
2
Cl CFC-114 15 220
CF
2
ClCF
3
CFC-115 5 550
Hydrochlorouorocarbons (HCFCs)
CF
2
ClH HCFC-22 Chlorodiuoromethane 130 11.8
CH
3
CFCl
2
HCFC-141b 6 9.2
CH
3
CF
2
Cl HCFC-142b 2-Diuorochloroethane 8 18.5
Other chlorocarbons
CCl
4
Carbon tetrachloride 100 35
CH
3
CCl
3
Methyl chloroform 90 4.8
CH
3
Cl Methyl chloride 610 1.3
Other chlorinated compounds
HCl Hydrochloric acid 11000 -1
Bromocarbons
Halons
CF
3
Br H-1301 Triuorobromomethane 2 65
CF
2
ClBr H-1211 Diuorochlorobromomethane 2 11
CF
2
BrCF
2
Br H-2402 1-Diuorobromo- 1.5 2230
2-diuorobromoethane
Other bromocarbons
CH
3
Br Methyl bromide 12 0.7
Fluorocarbons and uorine compounds
Hydrouorocarbons (HFCs)
CH
2
FCF
3
HFC-134a 1-Fluoro-2-triuoroethane 4 13.6
Peruorocarbons (PFCs)
C
2
F
6
Peruoroethane 4 10 000
Other uorinated compounds
SF
6
Sulfur hexauoride 3.7 3200
Sources: Shen et al. (1995); Singh (1995); WMO (1998); Mauna Loa Data Center (2001).
ammable and toxic gas. CFC-12 and subsequent CFCs were inexpensive, non-
toxic, nonammable, nonexplosive, insoluble, and chemically unreactive under
tropospheric conditions; thus, they became popular.
In 1931, CFC-12 was produced by the DuPont chemical manufacturer under the
trade name Freon. Its rst use was in small ice cream cabinets. In 1934, it was used
in refrigerators and whole-room coolers. Soon after, it was used in household and
400
automotive air conditioning systems. In 1932, CFC-11 was rst produced for use
in large air conditioning units. CFCs became airborne only when coolants leaked
or were drained.
In 1943, Goodhue and Sullivan of the United States Department of Agricul-
ture developed a method to use CFC-11 and -12 as a propellant in spray cans.
CFCs owed out of a spray cans nozzle, carrying with them a mist containing
other ingredients. Spray cans were used to propel hair sprays, paints, deodorants,
disinfectants, polishes, and insecticides.
CFC-11 and 12 have also been used as blowing agents in foamproduction. Foam
is used in insulation, disposable cups and cartons, and re extinguishers. CFCs are
released to the air during foam-production, itself. CFCs in the air spaces of foam
are usually conned and not an important source of atmospheric CFCs.
Chlorouorocarbons are a subset of chlorocarbons, which are compounds con-
taining carbon and chlorine. Hydrochlorouorocarbons (HCFCs) are another sub-
set of chlorocarbons. HCFCs are similar to CFCs, except that HCFCs have at least
one hydrogen atom. The hydrogen atom allows HCFCs to be broken down in the
troposphere by reaction with OH, a chemical that does not readily break down
CFCs. Because HCFCs break down more readily than do CFCs, a smaller percent-
age of emitted HCFCs than CFCs reaches the stratosphere. Nevertheless, because
HCFCs contain chlorine and some HCFCs reach the stratosphere, HCFCs are still a
danger to stratospheric ozone. HCFC-22, rst produced in 1943, is the most abun-
dant HCFC in the air today. HCFC-22 has been used as a refrigerant, spray-can
propellant, and blowing agent in foam production.
Other chlorocarbons include carbon tetrachloride (CCl
4
), methyl chloroform
(CH
3
CCl
3
), and methyl chloride (CH
3
Cl). Carbon t etrachloride is used as an in
termediate in the production of CFCs and HCFCs, and as a solvent and grain fumi-
gant. Methyl chloroform is used as a degreasing agent, a dry-cleaning solvent, and
an industrial solvent. Methyl chloride is produced synthetically only in small quan-
tities and used in the production of silicones and tetramethyl lead intermediates
(Singh 1995). Most methyl chloride in the air is produced biogenically in the oceans.
Another chlorine-containing gas in the troposphere is hydrochloric acid (HCl).
HCl has larger natural than anthropogenic sources. Natural sources include evap-
oration of chloride from sea-spray and volcanic emission.
Although chlorine-containing compounds are more abundant than are bromine-
containing compounds, the latter compounds are more efcient, molecule for
molecule, at destroying ozone. The primary source of stratospheric bromine is
methyl bromide (CH
3
Br), which is produced biogenically in the oceans and emitted
as a soil fumigant. Other sources of bromine are a group of synthetically produced
compounds termed Halons, which are used in re extinguishers and as fumigants.
The most common Halons are H-1301 (CF
3
Br), H-1211 (CF
2
ClBr) and H-2402
(CF
2
BrCF
2
Br). Methyl bromide and Halons are bromocarbons because they con-
tain both bromine and carbon.
Compounds that contain hydrogen, uorine, and carbon but not chlorine or
bromine are hydrouorocarbons (HFCs). HFCs were produced in abundance only
recently as a replacement for CFCs and HCFCs. Because the uorine in HFCs has
401
0 100 200 300 400 500 600
0
10
20
30
40
50
A
l
t
i
t
u
d
e

(
k
m
)
CFC-11
HCFC-22
CFC-12
C
C
l
4
(
g
)
Mixing ratio (pptv)
Tropopause
Figure 11.7 Variation of CFC-11,
CFC-12, HCFC-22, and CCl
4
(g) with
altitude at 30
N latitude. Smoothed
and scaled from Jackman et al. (1996)
to present-day near-surface mixing
ratios.
little effect on ozone, production of HFCs may increase in the future. Unfortu-
nately, because they absorb thermal-infrared radiation, HFCs will enhance global
warming if their use increases. The most abundantly emitted HFC to date has been
HFC-134a (CH
2
FCF
3
(g)). Related to HFCs are peruorocarbons (PFCs), such as
peruoroethane (C
2
F
6
), and sulfur hexauoride (SF
6
).
11.3.5.2 Stratospheric breakdown of chlorinated compounds
Once emitted, CFCs take about one year to mix up to the tropopause. Because
they are chemically unreactive, and ultraviolet wavelengths that reach the tropo-
sphere are too weak to break them down, CFCs are not removed chemically from
the troposphere. Instead, they become well mixed there. Today, the tropospheric
mixing ratios of CFC-11 and CFC-12, the two most abundant CFCs, are about
270 and 550 pptv, respectively (Table 11.8 and Fig. 11.7). CFCs break down only
when they reach the stratosphere.
Because the stratosphere is one large temperature inversion, vertical transport
of ozone to the stratosphere is slow. About 10 Mt of chlorine in the form of
CFCs reside in the troposphere, and the transfer rate of CFC-chlorine from the
troposphere to the middle stratosphere is about 0.1 Mt per year. In this simpli-
ed scenario, the average time required for a CFC molecule to migrate from the
troposphere to the middle stratosphere is about 100 years.
CFCs break down only when they reach the stratosphere, where they are exposed
to far-UV radiation (wavelengths of 0.01 to 0.25 m). This exposure occurs at an
altitude of 12 to 20 kmand higher. At such altitudes, far-UVwavelengths photolyze
402
CFC-11 and CFC-12 by
F C
Cl
Cl
Cl
+ h F C
Cl
Cl
+
Cl
< 250 nm
(11.126)
F C
Cl
Cl
F
+ h F C
Cl
F
+ Cl < 226 nm
(11.127)
decreasing CFC mixing ratios in the stratosphere, as seen in Fig. 11.7. At 25 km,
the e-folding lifetimes of CFC-11 and CFC-12 against photolysis under maximum-
sunlight conditions are on the order of 23 and 251 days, respectively. Average
lifetimes are on the order of two to three times these values. Thus, the time
required to break down CFCs in the middle stratosphere (on the order of a year)
is much less than the time required for CFCs to get there from the upper tro-
posphere (on the order of decades). Table 11.8 indicates that the overall life-
times of CFC-11 and CFC-12 between release at the surface and destruction in
the middle stratosphere are about 45 and 100 years, respectively. The lifetime of
CFC-12 is longer than that of CFC-11, partly because the former compound must
climb to a higher altitude in the stratosphere before breaking apart than must the
latter.
HCFC-22 has a lower emission rate and reacts faster with OHin the troposphere
than do CFC-11 or CFC-12, so HCFC-22s mixing ratio and overall lifetime are
lower than are those of CFC-11 or CFC-12. However, HCFC-22 photolyzes slower
than do CFC-11 or CFC-12, so once HCFC-22 reaches the middle stratosphere,
its mixing ratio does not decrease so rapidly with height as do mixing ratios of
CFC-11 or CFC-12, as seen in Fig. 11.7.
In sum, the limiting factor in CFC decomposition in the stratosphere is not
transport from the surface to the tropopause or photochemical breakdown in the
stratosphere, but transport fromthe tropopause to the middle stratosphere. Because
of their long overall lifetimes, some CFCs emitted in the 1930s through 1950s are
still present in the stratosphere. Those emitted today are likely to remain in the air
until the second half of the twenty-rst century.
11.3.5.3 Anthropogenic versus natural sources of chlorine
Of the chlorine compounds reaching the stratosphere, >80 percent is anthro-
pogenic, about 15 percent is natural methyl chloride (emitted biogenically from
the oceans), and about 3 percent is natural hydrochloric acid (emitted volcanically
and evaporated from sea spray) (WMO 1995).
403
Although the oceans emit a tremendous amount of methyl chloride, the e-folding
lifetime against its loss by
H
H
Cl
H
H C Cl
H
+OH
H
2
O
C
(11.128)
is only 1.5 years, so a relatively small portion of it reaches the stratosphere. In the
stratosphere, methyl chloride photolyzes by
H C
H
Cl
H
+ h H C
H
H
Cl < 220 nm + (11.129)
with an e-folding lifetime at 25 km of 2.4 yr.
Volcanos emit lots of HCl, but most of it is either removed by clouds and rain
or lost chemically before reaching the stratosphere (Lazrus et al. 1979; Pinto et al.
1989; Tabazadeh and Turco 1993a). HCl is extremely soluble in water, so any
contact with clouds or precipitation removes it from the gas phase immediately.
HCl also has a relatively short (1530 day) e-folding lifetime against reaction with
OH, so HCl that is not rained out is generally destroyed chemically. Because of the
short lifetime of HCl (-0.1 yr), most HCl in the stratosphere is not transported
there but produced chemically there by CFC by-products.
In sum, the main natural source of chlorine into the stratosphere is methyl
chloride, but the quantity of its emission into the stratosphere is much lower than
that of anthropogenic chlorine compounds.
11.3.5.4 Catalytic destruction of ozone by chlorine
Once released from CFC and non-CFC parent compounds in the stratosphere
by photolysis, chlorine atoms may react in a chlorine catalytic ozone destruction
cycle,
Cl +O
3
Cl
O+O
2
(11.130)
Cl
O+
Cl +O
2
(11.131)
+ O
3
2O
2
At midlatitudes, the chain length of this cycle increases from about 10 in the lower
stratosphere to about 1000 in the middle and upper stratosphere (Lary 1997). The
Cl + ClO, where ClO is chlorine monoxide, that takes part in catalytic ozone
destruction is called active chlorine.
404
The primary removal mechanisms of active chlorine from the catalytic cycle
are chemical reactions that produce chlorine reservoirs, which are gases that store
active chlorine and prevent it from taking part in catalytic cycles. The major chlo-
rine reservoirs in the stratosphere are hydrochloric acid (HCl) and chlorine nitrate
(ClONO
2
).
The HCl reservoir forms by
Cl +
CH
4
HCl +
CH
3
H
O
2
HCl +O
2
H
2
HCl +
H
H
2
O
2
HCl +H
O
2
(11.133)
and the ClONO
2
reservoir forms by
Cl O
+
NO
2
Cl O
N
O
O
Chlorine
monoxide
Chlorine
nitrate
M
+
(11.134)
At any time, about 1 percent of the chlorine in the stratosphere is in the form of
active chlorine. Most of the rest is in the form of a chlorine reservoir.
The HCl chlorine reservoir leaks slowly. HCl reproduces active chlorine by
photolysis, reaction with OH, and reaction with O. These reactions are
HCl +
H+
Cl - 220 nm
OH
Cl +H
2
O
Cl +
OH
(11.135)
The e-folding lifetime of HCl against photolysis is about 1.5 yr at 25 km. HCl also
diffuses back to the troposphere, where it can be absorbed by clouds.
The ClONO
2
reservoir leaks primarily by the photolysis reactions
Cl O
N
O
O
+
h Cl +
O N
O
O
Chlorine
nitrate
Nitrate radical
< 400 nm
+ +
(11.136)
which occurs with an e-folding lifetime of about 4.5 h at 25 km.
11.3.5.5 Catalytic destruction of ozone by bromine
Like chlorine, bromine affects stratospheric ozone. The primary sources of strato-
spheric bromine are methyl bromide (CH
3
Br) and Halons. The tropospheric mixing
ratios of the most common Halons, CF
2
ClBr (H-1211) and CF
3
Br (H-1301), are
both about 2 pptv, less than 1 percent of the mixing ratios of CFC-11 and -12.
405
Nevertheless, the efciency of ozone destruction by the bromine catalytic cycle is
greater than is that by the chlorine catalytic cycle.
Methyl bromide and Halons photolyze in the stratosphere to produce atomic
bromine. Photolysis of methyl bromide occurs above 20 km by
H
H
Br
H
+
h H
H
H
+ Br < 260 nm
C C
(11.137)
The e-folding lifetime of CH
3
Br against loss by this reaction is about 10 days at
25 km.
Once released, atomic bromine takes part in the bromine catalytic ozone destruc-
tion cycle,
Br +O
3
Br
O+O
2
(11.138)
Br
O+
Br +O
2
(11.139)
+ O
3
2O
2
where bromine monoxide (BrO) is the partner of Br in this cycle. The chain
length of this cycle increases from about 100 at 20 km to about 10
4
at 40
50 km (Lary 1997). The chain length of the bromine catalytic cycle is longer
than that of the chlorine catalytic cycle because Br is chemically removed more
slowly from the bromine cycle than Cl is removed from the chlorine cycle. In
addition, the bromine reservoirs break down more quickly than do the chlorine
reservoirs.
Atomic bromine is removed from its catalytic cycle primarily by
Br +
O
2
HBr +O
2
H
2
O
2
HBr +H
O
2
(11.141)
where HBr is hydrobromic acid. When BrO is removed, it forms bromine nitrate
(BrONO
2
) by
Br O
+ NO
2
Br O
N
O
O
Bromine
monoxide
Bromine
nitrate
M
+
(11.142)
406
50
100
150
200
250
300
0
5
10
15
20
25
30
1980 1985 1990 1995 2000
O
z
o
n
e

m
i
n
i
m
u
m

(
D
U
)
O
z
o
n
e
-
h
o
l
e

a
r
e
a

(
1
0
6
k
m
2
)
Year
Area of N. America
Area of Antarctic
continent
Figure 11.8 Minimum ozone column abundances
and areal extent of the ozone hole over the Antarc-
tic region from 1979 to 2000. Data from NASA
Goddard Space Flight Center. For comparison, the
area of the Antarctic is about 13 10
6
km
2
and
the area of North America is about 24 10
6
km
2
.
The HBr and BrONO
2
reservoirs leak through the respective reactions
HBr +
OH
Br +H
2
O (11.143)
Br O
N
O
O
h Br +
O N
O
O
Bromine
nitrate
Nitrate radical
< 390 nm
+
+
+
(11.144)
The e-folding lifetime of BrONO
2
against the photolysis reaction is about 10 min
at 25 km.
11.3.6 Antarctic ozone hole
Every September through November (Southern Hemisphere spring) since 1980,
the minimum ozone column abundance over the Antarctic has decreased below
its yearly average. Figure 11.8 shows that, in 2000, the lowest measured column
abundance over the Antarctic was about 94 DU(occurring on September 29, 2000),
which was 68 percent less than 293.4 DU, the globally and year-2000 averaged
ozone column abundance. Between 1981 and 2000, the area over which ozone
depletes (the Antarctic ozone hole), increased. The ozone hole area is dened as
the area of the globe over which the ozone column abundance decreases below
220 DUs. The ozone hole in 2000 covered nearly 30 10
6
km
2
, an area larger than
the size of North America. Antarctic ozone depletion occurs between 12 and 24 km
in altitude.
407
During the Northern Hemisphere late winter and spring (MarchMay), an Arctic
ozone dent (reduction in ozone to 240 to 260 DU), smaller in magnitude than the
Antarctic ozone hole, appears. The hole and dent are caused by a set of interlinked
factors. One factor linking global ozone reductions to polar ozone depletion is the
presence of chlorine and bromine in the stratosphere.
11.3.6.1 Polar stratospheric cloud formation
The ozone hole over the Antarctic appears in part because the Antarctic winter
(JuneSeptember) is very cold. Temperatures are low because much of the polar
region is exposed to 24 hours of darkness each day during the winter, and a wind
system, the polar vortex, circles the Antarctic. The vortex is a polar-front jet-stream
wind system that ows around the Antarctic continent, trapping cold air within
the polar region and preventing an inux of warm air from outside this region.
Because temperatures are lowin the Antarctic stratosphere, optically thin clouds,
called polar stratospheric clouds (PSCs) form. These clouds have few particles
per unit volume of air in comparison with tropospheric clouds. Two major types
of clouds form. When temperatures drop to below about 195 K, nitric acid and
water vapor growon small sulfuric acidwater aerosol particles (Toon et al. 1986).
Initially, it was thought that nitric acid and water molecules deposited to the ice
phase in the ratio 1:3. Such ice crystals have the composition HNO
3
3H
2
Oand are
called nitric acid trihydrate (NAT) crystals. More recently, it was found that these
particles contain a variety of phases. Some contain nitric acid dihydrate (NAD)
(Worsnop et al. 1993), and others contain supercooled liquid water (liquid water
present at temperatures below the freezing point of water), sulfuric acid, and nitric
acid (Tabazadeh et al. 1994). Together, nitrate containing cloud particles that form
at temperatures belowabout 195 Kin the winter polar stratosphere are called Type
I polar stratospheric clouds.
When temperatures drop below the frost point of water, which is about 187 K
under typical polar stratospheric conditions, a second type of cloud forms. These
clouds contain pure water ice and are Type II polar stratospheric clouds. Usually,
about 90 percent of PSCs are Type I and 10 percent are Type II (Turco et al.
1989). Typical diameters and number concentrations of a Type I PSC are 1 m
and 1 particle cm
3
, respectively, although diameters vary from 0.01 to 3 m.
Typical diameters and number concentrations of a Type II PSC are 20 m and
0.1 particle cm
3
, respectively, although diameters vary from 1 to 100 m.
11.3.6.2 PSC surface reactions
Once PSCparticles form, chemical reactions take place on their surfaces. Reactions
involving a gas reacting on a particle surface are heterogeneous reactions. In the
present case, the surfaces are those of frozen particles, and the reactions occur after
a gas has diffused to and adsorbed to the particle surface. Adsorption is the process
by which a gas collides with and bonds to a liquid or solid surface. Adsorption
differs from absorption, which is a process by which a gas penetrates into the inner
structure of a liquid or solid. One type of adsorption bonding is iondipole bonding,
408
which occurs when the surface is charged and the gas is polar. A second type is
dipoledipole bonding, which occurs when the surface and gas are polar. Once a
gas molecule adsorbs to a surface, it can desorb (break away) from the surface.
If it stays adsorbed to the surface, it can diffuse to another site on the surface.
During the diffusion process, the adsorbed gas may collide and chemically react
with another adsorbed molecule. Agas molecule suspended above a particle surface
may also collide and react with the adsorbed molecule. In both cases, an adsorbed
product is formed. This product can either desorb from the surface, diffuse on the
surface, or participate in additional chemical reactions.
The primary heterogeneous reactions that occur on Types I and II PSC surfaces
are
ClONO
2
(g) +H
2
O(a) HOCl(g) +HNO
3
(a) (11.145)
ClONO
2
(g) +HCl(a) Cl
2
(g) +HNO
3
(a) (11.146)
N
2
O
5
(g) +H
2
O(a) 2HNO
3
(a) (11.147)
N
2
O
5
(g) +HCl(a) ClNO
2
(g) +HNO
3
(a) (11.148)
HOCl(g) +HCl(a) Cl
2
(g) +H
2
O(a) (11.149)
In these reactions, (g) denotes a gas and (a) denotes an adsorbed species. Additional
reactions exist for bromine. Laboratory studies show that HCl readily coats the
surfaces of Types I and II PSCs. When ClONO
2
, N
2
O
5
, or HOCl impinges upon
the surface of a Type I or II PSC, it can react with H
2
O or HCl already on the
surface. The products of these reactions are adsorbed species, some of which stay
adsorbed, whereas others desorb to the vapor phase.
In sum, heterogeneous reactions convert relatively inactive forms of chlorine
in the stratosphere, such as HCl and ClONO
2
, to photochemically active forms,
such as Cl
2
, HOCl, and ClNO
2
. This conversion process is chlorine activation.
The most important heterogeneous reaction is (11.146) (Solomon et al. 1986;
McElroy et al. 1986), which generates gas-phase molecular chlorine, Cl
2
. Reac-
tion (11.147) does not activate chlorine. Its only effect is to remove nitric acid
from the gas phase. When nitric acid adsorbs to a Type II PSC, which is larger than
a Type I PSC, the nitric acid can sediment out along with the PSC to lower regions
of the stratosphere. This removal process is stratospheric denitrication. Denitri-
cation is important because it removes nitrogen that might otherwise reform Type
I PSCs or tie up active chlorine as ClONO
2
.
11.3.6.3 Reaction probabilities
Rate coefcients for (11.145)(11.149) have the form
k
s.q
=
1
4
v
q
q
a (11.150)
409
Table 11.9 Estimated reaction probabilities for the gases in
the reactions (11.145)(11.149) on Types I and II PSC
surfaces
Reaction probability
Reaction Type I PSCs Type II PSCs
ClONO
2
(g) + H
2
O(a) 0.001 0.3
ClONO
2
(g) + HCl(a) 0.1 0.3
N
2
O
5
(g) + H
2
O(a) 0.0003 0.01
N
2
O
5
(g) + HCl(a) 0.003 0.03
HOCl(g) + HCl(a) 0.1 0.3
Source: DeMore et al. (1997) and references therein.
where v
q
is the thermal speed of the impinging gas (cms
1
),
q
is the reaction prob-
ability (also called the uptake coefcient) of the gas (dimensionless), and a is the
surface-area concentration (square centimeters of surface per cubic centimeter of
air) of all particles on which reactions occur. Equation (11.150) implicitly includes
the concentration of the adsorbed reactant but not of the gas reactant; thus, it is a
pseudo-rst-order rate coefcient (s
1
). When multiplied by the gas-phase reactant
concentration, (11.150) gives a rate (molec. cm
3
s
1
).
The average thermal speed of a gas from (2.3) is
v
q
=
8k
B
T

M
q
(11.151)
where

M
q
= m
q
A is the mass (g) of one gas molecule. Example 2.1 gives the
average thermal speed of air molecules from this equation.
A reaction probability is the laboratory-measured fractional loss of a species
from the gas phase due to reaction with a particle surface, and it takes into account
diffusion of the gas to the surface as well as reaction with the surface. Estimated
reaction probabilities for the gases in (11.145)(11.149) on Types I and II PSCs
are given in Table 11.9.
Reaction probabilities can be derived by considering adsorption, desorption,
and reaction on a particle surface (e.g., Tabazadeh and Turco 1993b; Adamson
1990). The rate of change of species concentration in the rst molecular layer over
a surface is
dn
s.q
dt
= k
a.q
p
q
(n
m
n
s.q
) k
d.q
n
s.q
k
s.q
n
s.q
(11.152)
where n
s.q
is the surface-area concentration of adsorbed gas q (molec. cm
2
), k
a.q
is the rate at which gas molecules adsorb to the surface (molec. hPa
1
s
1
), p
q
is the partial pressure of the gas over the surface (hPa), n
m
10
15
sites cm
2
is
approximately the number concentration of adsorption sites on the surface, k
d.q
410
is the rate at which adsorbed molecules desorb from the surface (s
1
), and k
s.q
is
the rst-order rate coefcient for the loss of adsorbed species q due to chemical
reaction with another adsorbed species or with a gas (s
1
). Equation (11.152) states
that the change in concentration of molecules adsorbed to the surface of a particle
equals an adsorption ux minus a desorption ux minus a reaction ux.
The rate of molecular adsorption is
k
a.q
=

q
0
2

M
q
k
B
T
exp
E
a.q
+
f.q
R
(11.153)
where
q
is the mass accommodation coefcient of the adsorbing gas,
0
10
15
cm
2
is the molecular surface area of one site,

M
q
is the mass (g) of one adsorbed
molecule, E
a.q
is the activation energy of adsorption (J mol
1
), is a factor that
allows E
a.q
to change with surface coverage, and
f.q
is fractional surface coverage.
The mass accommodation coefcient (sticking coefcient) is the fractional number
of collisions of gas q with a particle that results in the gas sticking to the particles
surface.
The rate of molecular desorption is
k
d.q
=
0
exp
LG
d.q
+ E
d.q
R
(11.154)
where
0
10
13
s
1
is the frequency of atomic vibrations on the surface, LG
d.q
is
the change in Gibbs free energy of desorption (J mol
1
), and E
d.q
is the activation
energy of desorption (J mol
1
). The free energy of desorption is related to the free
energy of adsorption and entropy of adsorption by
LG
d.q
= LG
a.q
= LH
a.q
(
f.q
) TLS
a.q
(
f.q
) (11.155)
where LH
a.q
(
f.q
) is the enthalpy of adsorption (J mol
1
), and LS
a.q
(
f.q
) is the
entropy of adsorption (J mol
1
K
1
), both of which depend on the fractional
surface coverage. If the adsorbed species diffuses on the surface, the entropy of
adsorption is
LS
a.q
(
f.q
) = R
ln
f.q
+ R
ln(

M
q
T
0
) R
ln
T
52

M
32
q
+96.65 (11.156)
Surface coverage is the fraction of available surface sites lled with adsorbed
molecules. The steady-state surface coverage of an adsorbed substance is obtained
by setting (11.152) to zero. The result is
f.q
=
n
s.q
n
m
=
k
a.q
p
q
k
d.q
+k
s.q
+k
a.q
p
q
(11.157)
The adsorption minus desorption ux in (11.152) can be written in terms of a
net diffusive ux from the gas phase and a reaction probability:
1
4
v
q
q
N
q
= k
a.q
p
q
(n
m
n
s.q
) k
d.q
n
s.q
(11.158)
411
where N
q
= p
q
k
B
T is the number concentration of gas molecules suspended over
the particle surface (molec. cm
3
). Substituting (11.158) into (11.152) gives
dn
s.q
dt
=
1
4
v
q
q
N
q
k
s.q
n
s.q
(11.159)
Setting (11.159) equal to zero and solving for the steady-state reaction probability
gives
q
=
4k
s.q
n
s.q
v
q
N
q
=
4k
s.q
f.q
n
m
v
q
N
q
(11.160)
Parameters required for
f.q
, such as E
a.q
, E
d.q
, , and k
s.q
, are difcult to obtain.
Thus, laboratory-derived reaction probabilities are generally used and substituted
into (11.150) to approximate k
s.q
.
11.3.6.4 Springtime polar chemistry
Chlorine activation occurs during the winter over the polar stratosphere. When
the Sun appears over the horizon in the early spring, Cl-containing gases created
by PSC reactions photolyze by
Cl
2
+h 2
Cl - 450 nm (11.161)
HOCl +h
Cl +
OH - 375 nm (11.162)
ClNO
2
+h
Cl +
NO
2
- 372 nm (11.163)
Once Cl has been released, it attacks ozone. The catalytic cycles that destroy ozone
in the springtime polar stratosphere differ from that shown in (11.130)(11.132),
which reduces ozone on a global scale. One polar stratosphere catalytic ozone
destruction cycle is
2 (
Cl +O
3
Cl
O+O
2
) (11.164)
Cl
O+Cl
O
M
Cl
2
O
2
(11.165)
Cl
2
O
2
+h ClO
O+
Cl - 360 nm (11.166)
ClO
O
M
Cl +O
2
(11.167)
2O
3
3O
2
(11.168)
where Cl
2
O
2
is dichlorine dioxide and ClOO is a chlorine peroxy radical. This
mechanism, called the dimer mechanism (Molina and Molina 1986), is important
412
62%
HCl
37%
ClONO
2
1% Cl, ClO
Before PSC and photolysis
reactions
HCl
ClONO
2
Cl, ClO
After PSC and photolysis
reactions
Figure 11.9 Pie chart showing conversion of chlorine
reservoirs to active chlorine. During chlorine activation
on PSCs, HCl and ClONO
2
are converted to potentially
active forms of chlorine that are broken down by sun-
light in springtime to form Cl. Cl forms ClO, both of
which react catalytically to destroy ozone.
in the springtime polar stratosphere because, at that location and time, the ClO
required for (11.165) is concentrated enough for the reaction to proceed rapidly.
A second cycle is
Cl +O
3
Cl
O+O
2
(11.169)
Br +O
3
Br
O+O
2
(11.170)
Br
O+Cl
Br +
Cl +O
2
(11.171)
2O
3
3O
2
(11.172)
(McElroy et al. 1986), which is important in the polar lower stratosphere. In sum,
chlorine activation and springtime photochemical reactions convert chlorine from
reservoir forms, such as HCl and ClONO
2
, to active forms, such as Cl and ClO, as
shown in Fig. 11.9. The active forms of chlorine destroy ozone in catalytic cycles.
Every November, the Antarctic warms up sufciently for the polar vortex to
break down and PSCs to melt, evaporate, and sublimate. Ozone from outside
the polar region advects into the region. Ozone also regenerates chemically, and
chlorine reservoirs of ClONO
2
and HCl reestablish themselves. Thus, the Antarctic
ozone hole is an annual, regional phenomenon that is controlled primarily by the
temperature of the polar stratosphere and the presence of chlorine and bromine.
The radial extent of the hole has grown and minimumozone values have decreased
steadily over the past decades. The ozone dent over the Arctic is not nearly so large
or regular as is that over the Antarctic because the vortex around the Arctic is
much weaker and temperatures over the Arctic do not drop so low as they do
over the Antarctic. Thus, PSC formation and subsequent chemical reaction are less
widespread over the Arctic than over the Antarctic.
413
0
50
100
150
200
250
300
350
0 100 200 300 400 A
v
e
r
a
g
e

g
l
o
b
a
l

o
z
o
n
e

c
o
l
u
m
n
(
D
o
b
s
o
n

u
n
i
t
s
)
Day and date of simulation
10/1 1/7 4/17 7/26
11/4
No chlorine
With chlorine
Figure 11.10 Change in ozone column abundance,
averaged over the globe, during two global model
simulations in which chlorine was present and
absent. In both cases, ozone was initially removed
from the model atmosphere on October 1, 1988.
Bromine was not included in either simulation. In
the chlorine case, the column abundance regenerated
to a maximum of 300.4 DU, which compares with
an observed maximum in 1989 of about 300 DU.
In the no-chlorine case, the ozone layer regenerated
to a maximum of 308.5 DU, which compares with
an observed average abundance between 1964 and
1980 of 306.4 DU (Bojkov and Fioletov 1995).
11.3.7 Recovery of stratospheric ozone
In September 1987, an international agreement, the Montreal Protocol, was signed
limiting the production of CFCs and Halons. The Montreal Protocol has since
been modied several times to accelerate the phase-out of CFCs. The major effect
of the Montreal Protocol and later amendments has been nearly to eliminate the
emission of CFCs. The reduction in CFC emission will, in the long term, allow
the stratospheric ozone layer to recover, but because of the long lifetimes of CFCs
the recovery may take decades.
However, once CFCs and the active chlorine they produce are removed from
the stratosphere, the ozone layer should recover rapidly. Suppose that all ozone in
the stratosphere were destroyed, all ozone-destroying compounds were removed,
but all oxygen remained. How long would the ozone layer take to regenerate? An
estimate can be obtained from Fig. 11.10, which shows results from two global-
model simulations of the atmosphere in which all ozone was initially removed, but
oxygen was not. In the rst simulation, ozone regeneration was simulated in the
absence of chlorine and bromine. In the second, ozone regeneration was simulated
in the presence of 1989 concentrations of chlorine, but in the absence of bromine.
In both simulations, the globally averaged column abundance of ozone regenerated
to relatively steady values in less than a year. Regeneration during the simulation in
which chlorine was initially present was about 2 to 3 percent less than that during
414
11.4 Summary
0 2 4 6 8 10
0
10
20
30
40
1 h
6 h
1 d
5 d
50 d
464 d
Ozone volume mixing ratio (ppmv)
A
l
t
i
t
u
d
e

(
k
m
)
0 10 20 30 40 50 60
0
10
20
30
40
1 h
6 h
1 d
5 d
50 d
464 d
Ozone (10
11
molecules cm
3
)
A
l
t
i
t
u
d
e

(
k
m
)
(a) (b)
Figure 11.11 Time evolution of the vertical prole of ozone (a) volume mix-
ing ratio and (b) number concentration at 34
N latitude, starting with zero

ozone on June 21, 1988, as predicted by a one-dimensional photochemical
radiative model. Mixing ratios regenerated starting fromthe top of the strato-
sphere. Number concentrations regenerated starting from the bottom.
the no-chlorine case, consistent with the estimated global reduction in ozone of 2
to 3 percent between the 1970s and 1989 due to chlorine-containing compounds.
Figures 11.11 (a) and (b) show the time-evolution of the vertical mixing ratio
and number concentration, respectively, of ozone from June 21, 1988 through
September 26, 1989 (464 days later) in a one-dimensional column calculation in
which all ozone was removed initially from the atmosphere. The gure shows that,
within a few hours after the start, ozone mixing ratio regenerated starting from
the top of the stratosphere, and number concentrations regenerated starting from
the bottom. Number concentration is an absolute quantity, whereas mixing ratio
is a relative quantity. In sum, not only does the total column abundance of ozone
regenerate quickly after removal of chlorine and bromine, but so does the vertical
prole of ozone.
11.4 SUMMARY
In this chapter, chemistry of the free troposphere, urban areas, the global strato-
sphere, and the polar stratosphere were discussed. Important chemical equations
and reaction pathways were given. The focus of many atmospheric pollution stud-
ies is on ozone formation and destruction. In urban regions, the problem is ozone
formation. In the stratosphere, the problem is ozone destruction. The chemical
pathways that form and destroy ozone in each region are different. In the free
troposphere, ozone is governed by NO reaction with O
3
and by NO
2
photolysis.
In urban regions, ozone is governed by NO reaction with ROGs and O
3
and NO
2
photolysis. Ozone production in the stratosphere is governed by O
2
photolysis.
Appendix Table B.4 summarizes most of the reactions described in this chapter. In
the next chapter, methods of solving chemical ordinary differential equations are
described.
415
Table 11.10 Mixing-ratio data from the Los Angeles basin for Problem 11.1
O
3
.m

O
3
.p
NO

NO
2
(ppmv) (ppmv)
Station (ppmv) (ppmv)
NO
2
NO modeled photostationary
O
3
.m
O
3
.p
O
3
.p
100%
8:30 a.m.
Anaheim 0.1757 0.0760 0.0058
Costa Mesa 0.0607 0.0446 0.0119
Hawthorne 0.0654 0.0525 0.0104
La Habra 0.0764 0.0501 0.0092
Lynwood 0.1209 0.0625 0.0073
Pasadena 0.1179 0.0740 0.0089
Reseda 0.0704 0.0645 0.0125
Simi Valley 0.0232 0.0355 0.0204
1:30 p.m.
Anaheim 0.0088 0.0374 0.0859
Costa Mesa 0.0047 0.0131 0.0617
Hawthorne 0.0169 0.0394 0.0471
La Habra 0.0108 0.0461 0.0879
Lynwood 0.0096 0.0372 0.0778
Pasadena 0.0321 0.0693 0.0491
Reseda 0.0091 0.0495 0.1216
Simi Valley 0.0031 0.0148 0.0990
11.5 PROBLEMS
11.1 (a) Fill in Table 11.10, assuming that the photolysis rates of NO
2
at 8:30 a.m.
and 1:30 p.m. were J = 0.008 s
1
and J = 0.01 s
1
, respectively, and T =
291 K at 8:30 a.m. and T = 298 K at 1:30 p.m. Assume p
d
= 1005 hPa in
both cases.
(b) Modeledozone mixing ratios inTable 11.10 were calculatedwitha chem-
ical model that included all the urban reactions in Appendix Table B.4,
including organic reactions. Explainthe difference betweenmodeledand
photostationary-state mixing ratios, if any, between the morning and the
afternoon. Why do morning andafternoonvalues differ less at Costa Mesa
than at other locations?
11.2 (a) Calculate the thermal speed of chlorine nitrate when T = 192 K. At this
temperature, what type of PSC might be present?
(b) Calculate the pseudo-rst-order rate coefcient of chlorine nitrate on the
surfaces of these cloud particles, assuming the surface is coated with
HCl. (Hint: Use a typical number concentration and diameter, given in
the text, of the chosen PSC type.)
(c) If
ClONO
2
= 1.0 ppbv initially, and only surface chemistry is considered,
how much chlorine nitrate will remain after 15 days due to the reaction
from Problem 11.2(b). What is the resulting mixing ratio of Cl
2
if none
existed initially?
11.3 Write the rate expressions for the reactions (11.164)(11.168), and write the
rst-derivative expressions for each species (except for O
2
and M) in the
reactions (ignore the factor of two in the rst reaction). Assume M is already
416
included in the rate coefcients. What is the steady-state expression for the
Cl concentration from the rate expression for Cl? If T = 197 K and p
d
=
25 hPa, calculate the rate coefcient of each of the kinetic reactions.
11.4 Write out all the reactions in Appendix Table B.4 that destroy molecu-
lar hydrogen (H
2
). Calculate the e-folding lifetime against loss in all cases
assuming T = 290 K, [OH] = 5.5 10
5
molec. cm
3
, [O(
1
D)] = 1.0 10
3
molec. cm
3
, and [Cl] = 1.0 10
3
molec. cm
3
. Based on the result, which
of these is the most important reaction destroying molecular hydrogen in
the atmosphere?
11.5 (a) Input all the inorganic kinetic reactions and either the (i) chlorine,
(ii) bromine, (iii) sulfur, or (iv) isoprene kinetic reactions from Appendix
Table B.4 into the computer le developed for Problem 10.7. If the iso-
prene mechanismis chosen, many nonaromatic organic reactions should
also be included to account for the chemistry of isoprene by-products.
Calculate rate coefcients for each reaction. For the chlorine and bromine
cases, assume T = 192 K and p
d
= 25 hPa. For the isoprene and sulfur
cases, assume T = 298.15 K and p
d
= 1013 hPa.
(b) Input the necessary photolysis reactions from Appendix Table B.4 to
complete the mechanism chosen in part (a). If a stratospheric mecha-
nism was chosen, input the stratospheric (25 km) peak photolysis rate
coefcients from the table into the le. Otherwise, input the surface
(0 km) peak rate coefcients. Write a script to scale the peak rate coef-
cients to any time of day. Assume the coefcients are zero before 6:00
a.m. and after 6:00 p.m. but vary as a sine function from 6:00 a.m. to 6:00
p.m., peaking at noon. Calculate rate coefcients for 9:00 a.m. and 5:00
p.m.
417
12
Methods of solving chemical ordinary
S
everal methods have been developed to solve chemical ordinary differ-
ential equations (ODEs). Those discussed in this chapter include analytical,
Taylor series, forward Euler, backward Euler, simple exponential, quasi-steady-
state, multistep implicitexplicit, backward differentiation (Gear), and family
methods. Additional methods include hybrid predictorcorrector methods (e.g.,
Young and Boris 1977), parameterization methods (e.g., Jacob et al. 1989b),
RungeKuttaRosenbrock schemes (Hairer and Wanner 1991; Sandu et al. 1997),
iterative backward Euler methods (e.g., Curtiss and Hirschfelder 1952; Shimazaki
and Laird 1970; Rosenbaum 1976; Hertel et al. 1993; Huang and Chang 2001),
hybrid NewtonRaphson iterative schemes (Gong and Cho 1993), GaussSeidel
methods (Verwer 1994), extrapolation techniques (Dabdub and Seinfeld 1995),
and projection methods (Sandu 2001) among others. The choice of a method for
solving atmospheric chemical problems depends on several factors, including stabil-
ity, accuracy, mass conservation, positivity, and speed. These factors are discussed
with respect to different solvers. Techniques of optimizing Gears method over a
three-dimensional grid are also described.
12.1 CHARACTERISTICS OF CHEMICAL ODES
Gas-phase chemical reactions are described by rst-order, rst-degree, homoge-
neous ordinary differential equations. Sets of gas-phase reactions are stiff in that
the chemical e-folding lifetimes of individual gases vary by many orders of magni-
tude. Because sets of chemical ODEs are stiff, some classical numerical methods are
not useful for solving them. For example, the fourth-order RungeKutta method
(Section 6.4.4.9) and the Richardson extrapolation/BulirschStoer method (Stoer
and Bulirsch 1980) are explicit methods that result in inefcient (slow) solutions
to stiff ODE problems. With an explicit method, nal concentrations at time t are
obtained by evaluating derivatives at the beginning of the current and previous
time steps (e.g., at times t h, t 2h, . . .), where a time step (s) h is the difference
between the current time (t) and the time of the previous time step (t h). When
an explicit technique is used to solve a stiff set of equations, the time step (h) is
limited by the e-folding lifetime of the shortest-lived chemical. This lifetime may
be 10
6
s or less. Time steps longer than the lifetime of the shortest-lived chemical
may destabilize the solution scheme. When a time step is always small, integration
418
12.1 Characteristics of chemical ODEs
of many reactions over days to months and over a large three-dimensional grid
requires a signicant amount of computational power and is often impractical.
Efcient solvers of stiff ODEs are semiimplicit in that their solutions at current
time t depend on derivatives evaluated at the current time, the beginning of the
current time step, and/or the beginning of previous time steps (times t, t h,
t 2h, . . .). Semiimplicit solvers can take time steps much longer than the e-folding
lifetime of the shortest-lived species and remain stable. Some accurate semiimplicit
schemes used for solving stiff ODEs are Gears method (Gear 1971), RungeKutta
Rosenbrock schemes (Kaps and Rentrop 1979; Hairer and Wanner 1991; Press
et al. 1992), and semiimplicit BulirschStoer schemes (Bader and Deuhard 1983;
Press et al. 1992), among others.
12.1.1 Initial value problems
Problems requiring the use of chemical ordinary differential equations are initial
value problems whereby the initial concentration of each species is known at time
t = 0, and a solution is desired at a nal time, t
f
> 0. Solutions are found by
integrating a set of chemical ODEs one time step (h) at a time between t = 0 and
t = t
f
.
In the case of gas chemistry, concentrations are denoted by N (molec. cm
3
).
At any given time t, the concentration of a species i is N
i,t
, and the vector of
concentrations of a set of K species is
N
t
= [N
1.t
. N
2.t
. . . . . N
i.t
. . . . . N
K.t
] (12.1)
Concentrations of an individual species and a set of species one time step backward
are N
i.th
and

N
th
, respectively. At the beginning of the rst time step, concentra-
tions of all species are set to initial concentrations. Thus,
N
i.th
= N
i.0
for i = 1. . . . . K (12.2)
where N
i.0
is the initial number concentration of gas i.
During a time step, an ODE scheme solves for all N
i.t
. Once values at time t
are found, t is replaced with t h and a new time step is solved for. With some
techniques, solutions depend on concentrations from several time steps backward
(times t h, t 2h. . . .). If the technique relies on values from two time steps
backward, the rst time step is solved for with a technique that depends on only
one time step backward, and subsequent time steps are solved for with the scheme
that depends on two time steps backward.
12.1.2 Properties of ODE solvers
For a chemical ODE solution scheme to be useful, it must be stable, accurate,
mass-conserving, positive denite, and computationally fast. In Section 6.4.1, a
numerical scheme was dened to be stable if the absolute-value difference between
the numerical and the exact solution did not grow over time. This denition, and
the denitions of unconditional stability, conditional stability, and unconditional
419
Methods of solving chemical ordinary differential equations
instability, given in the same section, apply to ODE and PDE solution techniques
alike.
Solutions must be accurate as well as stable. A method of testing a solvers
accuracy is to compare the time-dependent solution from it with an exact solution.
A normalized gross error (NGE) is then calculated as
NGE =
1
N
tim
N
tim
j =1
1
K
s.t
j
K
s.t
j
i =1
N
i.t
j
E
i.t
j
E
i.t
j
100% (12.3)
where N
tim
is the number of time steps, N
i.t
j
and E
i.t
j
are predicted and exact
concentrations, respectively, of species i at time t
j
, and K
s.t
j
is the number of exact
concentrations above a minimum cutoff concentration at time t
j
. Typical cutoff
concentrations for atmospheric gas chemistry range from10
3
to 10
3
molec. cm
3
.
In (12.3), concentrations may be compared at evenly spaced time intervals, such
as every one-half hour, instead of every time step. Exact solutions are obtained
by solving the equations with an integrator of known high accuracy or with the
given scheme using an extremely small time step. Good solvers of chemical ODEs
produce NGEs less than 1 percent.
A solver of chemical ODEs should be mass-conserving. A scheme is mass-
conserving if the mass of each element (e.g., N, O, H, or C) summed over all
species at the beginning of a simulation equals the mass of the element summed
over all species at the end of the simulation, provided no external sources or sinks
exist. Mass conservation cannot occur if individual reactions are not written in
a mass-conserving manner. The reaction NO + O
3
NO
2
+ O
2
is written in a
mass-conserving manner, but the reaction NO+ O
3
O
2
is not. If all reactions are
mass-conserving, the ODE solution scheme can be mass-conserving. Explicit ODE
solution schemes are by nature mass-conserving, since the addition of mass to one
species is accompanied by a loss of the same mass from another species. However,
explicit schemes may produce a concentration less than zero, in which case the
concentration is often set to zero or a small number, causing the scheme to gain
mass. In addition, since an inaccurate explicit scheme can still be mass-conserving,
mass conservation does not imply accuracy. Implicit or semiimplicit schemes have
the potential to be non-mass-conserving. It is more important for a scheme to be
accurate than exactly mass-conserving.
Concentrations predicted by chemical ODE solvers must exceed or equal zero.
This should not be a surprise, since in the atmosphere, gas concentrations exceed
or equal zero. If a scheme always predicts nonnegative concentrations, it is pos-
itive denite. An accurate chemical ODE solver is usually positive denite, since
correct solutions are always nonnegative. If a solver is unstable and/or inaccurate,
concentrations may fall below zero. If at least one concentration falls below zero
after a time step, the time step must either be re-solved with a shorter time step or
with a new solution method, or the concentration must be set to zero or above, in
which case mass is gained. A good solver of chemical ODEs can predict the time
step required to keep solutions stable and positive-denite.
420
12.3 Taylor series solution to ODEs
When used in atmospheric models, chemical ODE solvers should be fast as
well as accurate. All chemical ODE solvers are exactly accurate at a small
enough time step. A good solver can take a long step and maintain accuracy. For
three-dimensional modeling, accuracy with long steps is generally not enough.
Accurate solutions to chemical ODEs must be found over large model grids. In
the following sections, methods of solving chemical ODEs are discussed in light of
their stability, accuracy, mass conservation, positive deniteness, and/or speed.
12.2 ANALYTICAL SOLUTIONS TO ODES
The most accurate solution to a set of chemical ODEs is an analytical solution.
Analytical solutions to a single equation or a small set of equations are readily
found. Suppose nitrogen dioxide is lost by the photolysis reaction,
NO
2
+h
NO+
(J ) (12.4)
which has photolysis coefcient J (s
1
). The ODE describing this reaction is
d[NO
2
]
dt
= J [NO
2
] (12.5)
Integrating (12.5) gives its analytical solution at time t as
[NO
2
]
t
= [NO
2
]
th
e
J h
(12.6)
This equation states that the nitrogen dioxide concentration decays exponentially
if only photolysis is considered.
Example 12.1
For an initial concentration of [NO
2
]
th
= 10
10
molec. cm
3
and a photolysis
coefcient of 0.02 s
1
, (12.6) reduces to [NO
2
]
t
= 10
10
e
0.02h
, which is an exact
solution for the conditions given.
Whereas the solution to (12.5) was found easily, analytical solutions to a set of
more than a few equations are usually impractical to obtain. Thus, sets of chemical
ordinary differential equations are not solved analytically in atmospheric models.
12.3 TAYLOR SERIES SOLUTION TO ODES
Because analytical solutions are difcult to derive for chemical ODEs, numerical
solutions are needed. A useful method of solving sets of ODEs would appear to be
an explicit Taylor series expansion of species concentrations. In an explicit Taylor
series expansion, the concentration of species i at time t is approximated as
N
i.t
= N
i.th
+h
dN
i.th
dt
+
h
2
2
d
2
N
i.th
dt
2
+
h
3
6
d
3
N
i.th
dt
3
+ (12.7)
421
One difculty with using the Taylor series expansion is that expressions for higher
derivatives require combinations of many lower-derivative terms, increasing the
computational burden. Suppose three species NO, NO
2
, and O
3
are considered.
The explicit Taylor series expansion of the NO concentration is
[NO]
t
= [NO]
th
+h
d[NO]
th
dt
+
h
2
2
d
2
[NO]
th
dt
2
+
h
3
6
d
3
[NO]
th
dt
3
+ (12.8)
Similar equations can be written for O
3
and NO
2
. If one reaction,
NO+O
3

NO
2
+O
2
(12.9)
is considered, the rst, second, and third time derivatives of NO concentration
arising from this reaction are, respectively,
d[NO]
dt
= k
b
[NO] [O
3
] (12.10)
d
2
[NO]
dt
2
= k
b
d[NO]
dt
[O
3
] k
b
[NO]
d[O
3
]
dt
(12.11)
d
3
[NO]
dt
3
= k
b
d
2
[NO]
dt
2
[O
3
] 2k
b
d[NO]
dt
d[O
3
]
dt
k
b
[NO]
d
2
[O
3
]
dt
2
(12.12)
The rst, second, and third derivatives of O
3
and NO
2
are related to those of
NO by
d[O
3
]
dt
=
d[NO
2
]
dt
=
d[NO]
dt
(12.13)
d
2
[O
3
]
dt
2
=
d
2
[NO
2
]
dt
2
=
d
2
[NO]
dt
2
d
3
[O
3
]
dt
3
=
d
3
[NO
2
]
dt
3
=
d
3
[NO]
dt
3
respectively. Thus, to solve for changes in concentration due to one reaction only,
many terms are necessary. With the addition of tens or hundreds of reactions, the
number of terms increases.
A second problem with the explicit Taylor series method is that stiffness of the
system of equations prevents a stable and accurate solution unless the time step
used is small or the order of approximation is large. For these requirements to
be met, an explicit Taylor series expansion requires many short time steps, many
high-order terms, or both. In either case, the number of computations required to
maintain stability in a model with 10
3
10
6
grid cells and dozens of reactions is
prohibitively large.
12.4 FORWARD EULER SOLUTION TO ODES
If concentrations are approximated with the rst two terms of an explicit Taylor
series expansion, the approximation is rst order. The order of approximation
was dened in Section 6.4.1 as the order of the lowest-order term in a Taylor
series expansion of a derivative that is neglected when the approximation is made.
422
12.4 Forward Euler solution to ODEs
Dividing (12.7) by h and eliminating all terms of order h and higher gives a rst-
order approximation of an explicit Taylor series expansion, called the forward
Euler. The forward Euler solution for one species among a set of chemical ODEs
is
N
i.t
= N
i.th
+h
dN
i.th
dt
(12.14)
where the rst derivative is a function of N
i
and the concentrations of all other
species at time t h. In other words, dN
i.th
dt = f (N
th
).
Suppose the concentrations of NO, NO
2
, and O
3
are affected by the reactions,
NO+O
3

NO
2
+O
2
(k
1
) (12.15)
NO
2
+O
3

NO
3
+O
2
(k
2
) (12.16)
NO
2
+h
NO+
(J ) (12.17)
The explicit time derivative of NO
2
among these reactions is
d[NO
2
]
th
dt
= k
1
[NO]
th
[O
3
]
th
k
2
[NO
2
]
th
[O
3
]
th
J [NO
2
]
th
(12.18)
[NO
2
]
t
= [NO
2
]
th
+h(k
1
[NO]
th
[O
3
]
th
k
2
[NO
2
]
th
[O
3
]
th
J [NO
2
]
th
)
(12.19)
which is the forward Euler NO
2
concentration after one time step. This solution
can be generalized by dening total production and loss rates, respectively, of NO
2
as
P
c.NO
2
.th
= k
1
[NO]
th
[O
3
]
th
(12.20)
L
c.NO
2
.th
= k
2
[NO
2
]
th
[O
3
]
th
+ J [NO
2
]
th
(12.21)
Substituting these terms into (12.19) gives another form of the forward Euler solu-
tion,
[NO
2
]
t
= [NO
2
]
th
+h(P
c.NO
2
.th
L
c.NO
2
.th
) (12.22)
This notation can be generalized for any species i as
N
i.t
= N
i.th
+h(P
c.i.th
L
c.i.th
) (12.23)
An advantage of the forward Euler solution is that it requires minimal time for
computing production and loss terms. In addition, like the explicit Taylor series
expansion, it is exactly mass-conserving. For example, suppose reaction (12.15) is
solved alone. The forward Euler solutions for nitric oxide, ozone, nitrogen dioxide,
423
and oxygen are
[NO]
t
= [NO]
th
hk
1
[NO]
th
[O
3
]
th
(12.24)
[O
3
]
t
= [O
3
]
th
hk
1
[NO]
th
[O
3
]
th
(12.25)
[NO
2
]
t
= [NO
2
]
th
+hk
1
[NO]
th
[O
3
]
th
(12.26)
[O
2
]
t
= [O
2
]
th
+hk
1
[NO]
th
[O
3
]
th
(12.27)
respectively. For this set of four equations to conserve mass, the following relation-
ships must hold for nitrogen and oxygen, respectively:
([NO
2
] +[NO])
t
= ([NO
2
] +[NO])
th
(12.28)
(2[O
2
] +3[O
3
] +2[NO
2
] +[NO])
t
= (2[O
2
] +3[O
3
] +2[NO
2
] +[NO])
th
(12.29)
Substituting (12.24)(12.27) into these relationships gives exact equalities, indicat-
ing that the forward Euler conserves mass exactly.
The main disadvantage of the forward Euler is that it requires a small time step
to prevent concentrations of short-lived species from falling below zero. As such, it
requires many more time steps to complete a simulation than a semiimplicit scheme.
Because of the severity of this limitation, the forward Euler approximation is not
useful on its own to solve sets of stiff chemical equations.
12.5 BACKWARD EULER SOLUTION TO ODES
An approximation that avoids the time-step stability constraint of the forward
Euler solution is the linearized backward Euler solution. A linearized solution is
one in which concentrations of all species, except the species being solved for, are
set to concentrations fromprevious time steps (e.g., t h, t 2h, etc.) in derivative
terms. The concentration of the species of interest is set to time t and assumed to
be unknown when used in derivatives. In the backward Euler, only one time step
backward is considered. The linearized form of the backward Euler equation for a
species is
N
i.t
= N
i.th
+h
dN
i.t.th
dt
(12.30)
where the rst derivative is evaluated at time t h for all species except for the
species of interest, which is evaluated at time t. Thus, dN
i.t.th
dt = f (N
i.t
. N
th
).
The backward Euler solution can be illustrated with reactions (12.15)(12.17).
The linearized rst derivative of NO
2
from these reactions is
d[NO
2
]
t.th
dt
= k
1
[NO]
th
[O
3
]
th
k
2
[NO
2
]
t
[O
3
]
th
J [NO
2
]
t
(12.31)
where NO
2
is evaluated at time t. Substituting (12.31) into (12.30) gives
[NO
2
]
t
= [NO
2
]
th
+h(k
1
[NO]
th
[O
3
]
th
k
2
[NO
2
]
t
[O
3
]
th
J [NO
2
]
t
)
(12.32)
424
12.5 Backward Euler solution to ODEs
The production and loss terms in this equation are
P
c.NO
2
.th
= k
1
[NO]
th
[O
3
]
th
(12.33)
L
c.NO
2
.t.th
= k
2
[NO
2
]
t
[O
3
]
th
+ J [NO
2
]
t
(12.34)
respectively. All loss rates contain [NO
2
]
t
so the loss term can be divided by this
concentration to give an implicit loss coefcient,
A
c.NO
2
.th
=
L
c.NO
2
.t.th
[NO
2
]
t
= k
2
[O
3
]
th
+ J (12.35)
Substituting (12.33) and (12.35) back into (12.32) yields
[NO
2
]
t
= [NO
2
]
th
+h(P
c.NO
2
.th
A
c.NO
2
.th
[NO
2
]
t
) (12.36)
Since both sides of (12.36) contain [NO
2
]
t
, [NO
2
]
t
can be gathered on the left side
of the equation to produce
[NO
2
]
t
=
[NO
2
]
th
+hP
c.NO
2
.th
1 +hA
c.NO
2
.th
(12.37)
which is the linearized backward Euler solution. This solution can be generalized
for any species i with
N
i.t
=
N
i.th
+hP
c.i.th
1 +hA
c.i.th
(12.38)
The advantage of the linearized backward Euler solution is that the nal concen-
tration cannot fall below zero, regardless of the time step. As such, the scheme is
unconditionally stable. The disadvantage is that the method is not mass-conserving.
For example, suppose reaction (12.15) is solved alone. The linearized backward
Euler solutions for nitric oxide, ozone, nitrogen dioxide, and oxygen are
[NO]
t
= [NO]
th
hk
1
[NO]
t
[O
3
]
th
(12.39)
[O
3
]
t
= [O
3
]
th
hk
1
[NO]
th
[O
3
]
t
(12.40)
[NO
2
]
t
= [NO
2
]
th
+hk
1
[NO]
th
[O
3
]
th
(12.41)
[O
2
]
t
= [O
2
]
th
+hk
1
[NO]
th
[O
3
]
th
(12.42)
Substituting these equations into the nitrogen and oxygen mass-conservation rela-
tionships (12.28) and (12.29), respectively, gives inequalities, indicating that the
linearized backward Euler does not conserve mass. The reason is that reaction rates
are linearized differently when different species are solved, and for mass to be con-
served, reaction rates must be identical when different species are solved. Because
mass is not conserved, errors accumulate over time with the linearized backward
Euler.
425
12.6 SIMPLE EXPONENTIAL AND QUASI-STEADY-STATE
SOLUTIONS TO ODES
A method of solving ODEs, similar to the backward Euler, is the simple expo-
nential approximation. This solution is obtained by integrating a linearized rst
derivative. The advantage of the simple exponential approximation is that the nal
concentration cannot fall below zero; thus, the scheme is unconditionally stable.
The disadvantage is that, like the backward Euler approximation, the simple expo-
nential approximation is not mass-conserving.
The simple exponential solution is illustrated with the linearized rst derivative
from the backward Euler example. Writing (12.36) in differential form gives
d[NO
2
]
t
dt
= P
c.NO
2
.th
A
c.NO
2
.th
[NO
2
]
t
(12.43)
The integral of this equation is the simple exponential solution,
[NO
2
]
t
= [NO
2
]
th
e
hA
c.NO
2
.th
+
P
c.NO
2
.th
A
c.NO
2
.th
1 e
hA
c.NO
2
.th
(12.44)
which is written for any species i as
N
i.t
= N
i.th
e
hA
c.i.th
+
P
c.i.th
A
c.i.th
1 e
hA
c.i.th
(12.45)
When the implicit loss coefcient is zero (no loss), (12.45) simplies to N
i.t
=
N
i.th
+hP
c.i.th
. When the production termis zero, the solution simplies to N
i.t
=
N
i.th
e
A
c.i.th
. When the implicit loss coefcient is large (short-lived species), the
solution simplies to N
i.t
= P
c.i.th
A
c.i.th
, which is the steady-state solution to
the original ODE.
The simple exponential solution is similar to the linearized backward Euler solu-
tion, except that the former contains an analytical component and is slightly more
accurate than the latter. Because the simple exponential approximation requires
the computation of an exponent, it requires more computer time, per time step,
than does the backward Euler.
A method related to the simple exponential is the quasi-steady-state approxima-
tion (QSSA) method (e.g., Hesstvedt et al. 1978). With this method, the parameter
hA
c.i.th
is used to determine how a species concentration is solved during a time
step. When hA
c.i.th
- 0.01, the species lifetime is long, and its nal concentra-
tion is calculated with the forward Euler equation. When 0.01 hA
c.i.th
10,
the species lifetime is moderate, and its concentration is calculated with the sim-
ple exponential method. When hA
c.i.th
> 10, the species lifetime is short, and its
concentration is calculated with the steady state equation. In sum,
N
i.t
=
N
i.th
+h(P
c.i.th
L
c.i.th
) hA
c.i.th
- 0.01
N
i.th
e
hA
c.i.th
+
P
c.i.th
A
c.i.th
1 e
hA
c.i.th
0.01 hA
c.i.th
10
P
c.i.th
A
c.i.th
hA
c.i.th
> 10
(12.46)
426
12.7 Multistep implicitexplicit (MIE) solution to ODEs
Final concentrations at the end of a time step are found after several iterations
of (12.46). The iterative QSSA scheme can be accurate for many cases, but it is
mass-conserving only for long-lived species.
In sum, the forward Euler scheme is mass-conserving but positive-denite for
small time steps only; the backward Euler and simple exponential schemes are
always positive-denite, but not mass-conserving; and the iterative QSSA scheme
is often stable but generally not mass-conserving.
12.7 MULTISTEP IMPLICITEXPLICIT (MIE)
SOLUTION TO ODES
Apositive-denite, mass-conserving, unconditionally stable iterative technique that
takes advantage of the forward and backward Euler methods is the multistep
implicitexplicit (MIE) method (Jacobson 1994; Jacobson and Turco 1994). With
this method, concentrations are estimated with an iterated backward Euler cal-
culation, and the estimates are applied to reaction rates used in a forward Euler
calculation of nal concentrations. Upon iteration the forward Euler converges to
the backward Euler. Since backward Euler solutions are always positive, forward
Euler solutions must converge to positive values as well. A technique related to
the MIE method is merely iterating the backward Euler equation until convergence
occurs for all species. Such a method requires many more iterations than does the
MIE method.
The steps for determining nal concentrations with the MIE method are
described below. To illustrate, four species, NO, NO
2
, O
3
, and O, and two reac-
tions,
NO+O
3

NO
2
+O
2
(k
1
) (12.47)
O
3
+h O
2
+
(J ) (12.48)
are considered. The change in O
2
concentration by these reactions is ignored for
the illustration.
The rst step in the MIE solution is to initialize backward Euler concentrations
(molec. cm
3
) and maximumbackward Euler concentrations for each active species
i = 1, . . . , K with concentrations from the beginning of the simulation or, in the
case of a new time step, with concentrations from the end of the last time step.
Thus,
N
i.B.1
= N
i.th
(12.49)
N
i.MAX.1
= N
i.th
(12.50)
where the subscript B indicates a backward Euler concentration, and the subscript 1
indicates the rst iteration of a newtime step. In subsequent equations, the iteration
number is denoted by m. Here, m= 1. A maximum backward Euler concentration
is required to prevent backward Euler concentrations from blowing up to large
values upon iteration and is updated each iteration, as discussed shortly.
427
The second step is to estimate reaction rates by multiplying rate coefcients by
backward Euler concentrations. Examples of two-body and photolysis reaction
rates are
R
c.n.B.m
= k
n
N
i.B.m
N
j.B.m
R
c.n.B.m
= J
n
N
i.B.m
(12.51)
respectively, where R
c
is the rate of reaction (molec. cm
3
s
1
), n is the reaction-
rate number, m is the iteration number (= 1 for the rst iteration), k
n
is the kinetic
rate coefcient of the nth reaction, and J
n
is the photolysis rate coefcient of the
nth reaction. In the two-reaction example, the backward Euler rates are
R
c.1.B.m
= k
1
[NO]
B.m
[O
3
]
B.m
R
c.2.B.m
= J [O
3
]
B.m
(12.52)
respectively.
The third step is to estimate production rates, loss rates, and implicit loss coef-
cients for each species from the reaction rates just calculated. The backward Euler
production rate (molec. cm
3
s
1
) of species i is
P
c.i.B.m
=
N
prod.i
l=1
R
c.n
P
(l.i ).B.m
(12.53)
where N
prod.i
is the number of reactions in which species i is produced, and
R
c.n
P
(l.i ).B.m
is the lth backward Euler production rate of species i. The array n
P
(l. i )
gives the reaction number corresponding to the lth production term of species i. In
the two-reaction example, the summed production rates of O and NO
2
are
P
c.O.B.m
= J [O
3
]
B.m
P
c.NO
2
.B.m
= k
1
[NO]
B.m
[O
3
]
B.m
(12.54)
respectively, where the two active species produced have only one production term
each.
The backward Euler loss rate (molec. cm
3
s
1
) of a species is
L
c.i.B.m
=
N
loss.i
l=1
R
c.n
L
(l.i ).B.m
(12.55)
where N
loss.i
is the number of reactions in which species i is lost, and R
c.n
L
(l.i ).B.m
is the lth backward Euler loss rate of species i. The array n
L
(l. i ) gives the reaction
number corresponding to the lth loss termof species i. In the two-reaction example,
the summed loss rates of NO and O
3
are
L
c.NO.B.m
= k
1
[NO]
B.m
[O
3
]
B.m
L
c.O
3
.B.m
= k
1
[NO]
B.m
[O
3
]
B.m
+ J [O
3
]
B.m
(12.56)
The backward Euler implicit loss coefcient (s
1
) of a species is
A
c.i.B.m
=
L
c.i.B.m
N
i.B.m
(12.57)
In the two-reaction example, the implicit loss coefcients of NO and O
3
are
A
c.NO.B.m
= k
1
[O
3
]
B.m
A
c.O
3
.B.m
= k
1
[NO]
B.m
+ J (12.58)
428
Since a reactant concentration can equal zero, computing implicit loss coefcients
with (12.57) can result in a division by zero. If the coefcients are, instead, com-
puted directly from (12.58), division by zero is avoided.
The fourth step is to calculate backward Euler concentrations for all species at
iteration m + 1 with
N
i.B.m+1
=
N
i.th
+hP
c.i.B.m
1 +hA
c.i.B.m
(12.59)
Such estimates are used to calculate production and loss terms during the next
iteration.
The fth step is to calculate forward Euler concentrations for all species at
iteration m + 1 with
N
i.F.m+1
= N
i.th
+h(P
c.i.B.m
L
c.i.B.m
) (12.60)
The sixth step is to check convergence. Convergence is determined by rst check-
ing, at the end of each iteration during a time step, whether all forward Euler con-
centrations from (12.60) exceed or equal zero. If they do, a counter, n
P
, initialized
to zero before the rst iteration of the time step, is incremented by one and another
iteration is solved. If a single forward Euler concentration falls below zero during
a given iteration, the counter n
P
is reset to zero, and a new iteration is solved to
try to update n
P
again. Thus, whether the counter is updated or reset to zero is
determined by the following criteria:
N
F.m+1
0 n
P
= n
P
+1
N
i.F.m+1
- 0 n
P
= 0
(12.61)
where

N
F.m+1
is the entire set of forward Euler concentrations and N
i.F.m+1
is an
individual concentration. If all forward Euler concentrations 0 for N
P
iterations
in a row (e.g., if n
P
= N
P
), convergence is said to have occurred. N
P
is a constant
that depends on the number of reactions and their stiffness. Typical values are
5 for large sets of equations and 3050 for small sets of equations.
For faster solutions, the criteria can be modied so that, when hA
c.i.B.m
L
T
for a species during an iteration, the forward Euler concentration of the species
does not need to exceed zero for n
P
to avoid being reset to zero. In such cases, the
species is short-lived, and the backward Euler solution is more accurate than the
forward Euler solution. L
T
is a constant between 10
2
and 10
6
. Values of 10
2
speed
solutions but increase errors. Values of 10
6
slow solutions but decrease errors. In
sum, the following criteria identify conditions under which the counter is updated
or reset to zero under the modied convergence method:
N
F.m+1
0 for all species with hA
c.i.B.m
- L
T
n
P
= n
P
+1
N
i.F.m+1
- 0 and hA
c.i.B.m
- L
T
n
P
= 0 (12.62)
429
110
15
510
14
110
7
510
14
110
15
1000
10
5
10
7
10
9
10
11
10
13
10
15
1 10 100
N
O
2

a
n
d

O
3

(
m
o
l
e
c
u
l
e
s

c
m
3
)
O

(
m
o
l
e
c
u
l
e
s

c
m
3
)
Iteration number
O (FE)
O (BE)
NO
2
(FE)
NO
2
(BE)
O
3
(FE)
O
3
(BE)
Figure 12.1 Concentrations of three species iterated until conver-
gence with forward Euler (FE) and backward Euler (BE) calculations
during a MIE simulation where h = 10 s.
With the modied method, convergence is again obtained when n
P
= N
P
. When
the modied criterion is met, the nal concentrations are
N
i.t
=
N
i.B.m+1
backward Euler
hA
c.i.B.m
L
T
N
i.F.m+1
forward Euler
hA
c.i.B.m
- L
T
(12.63)
If the modied convergence criterion is not met, iterations continue.
Before iterations continue, a seventh step, to update maximum backward Euler
concentrations and to limit current backward Euler concentrations, is required.
Maximum backward Euler concentrations are limited to the larger of the current
backward Euler concentration and the initial concentration from the time step.
Current backward Euler concentrations are then limited to the smaller of their
current value and the maximum concentration from the last iteration. In other
words,
N
i.MAX.m+1
= max(N
i.B.m+1
. N
i.th
) (12.64)
N
i.B.m+1
= min(N
i.B.m+1
. N
i.MAX.m
) (12.65)
for each species. Note that the value of N
i.MAX
used in (12.65) is from the previous
iteration, while that calculated in (12.64) is from the current iteration. After these
updates, the iteration returns to (12.51) until convergence is obtained.
Here, it is shown that iterated backward Euler solutions converge to iterated
forward Euler solutions and positive numbers. At the end of any iteration, the
backward Euler concentration is
N
i.B.m+1
= N
i.th
+h(P
c.i.B.m
A
c.i.B.m
N
i.B.m+1
) (12.66)
which is simply (12.59) rearranged. Equation (12.66) must be positive, since
(12.59) cannot be negative. At the end of the same iteration, the forward Euler
yields
N
i.F.m+1
= N
i.th
+h(P
c.i.B.m
A
c.i.B.m
N
i.B.m
) (12.67)
430
110
0
510
11
110
12
1.510
12
210
12
2.510
12
310
12
0 4 8 12 16 20 24
C
o
n
c
e
n
t
r
a
t
i
o
n

(
m
o
l
e
c
u
l
e
s

c
m
3
)
Hours from start of simulation
HNO
3
O
3
Figure 12.2 Comparison of MIE solution (circles) with an
exact solution (lines) for two out of 92 species during a 24-h
simulation period. The time step taken with MIE was 10 s.
Photolysis rates changed every one-half hour during the rst
12 h and were zero during the second 12 h. MIE and exact
solutions are shown each one-half hour.
210
4
410
4
610
4
810
4
110
5
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
1 10 100 1000
N
o
.

o
f

i
t
e
r
a
t
i
o
n
s

r
e
q
u
i
r
e
d
A
v
e
r
a
g
e

e
r
r
o
r

p
e
r

s
p
e
c
i
e
s

(
%
)

Time step (s)
Figure 12.3 Effects of changing h with the MIE method when a
set of 92 ODEs were solved. Time steps of h = 2, 10, 100, 500,
and 1000 s are compared. The simulation time interval was 10 000
s. Small time steps resulted in a low normalized gross error after
10 000 s but required more iterations to complete the interval.
where A
c.i.B.m
N
i.B.m
= L
c.i.B.m
. For the forward and backward Euler solutions to
converge to each other,
N
i.B.m+1
= N
i.B.m
(12.68)
must be satised. This always occurs upon iteration of the backward Euler solution.
Thus, when the backward Euler solution converges, the forward Euler solution
must converge to the same value, and both must converge to positive numbers.
Figure 12.1 shows convergence of forward Euler to backward Euler concen-
trations versus iteration number for four species. Figure 12.2 compares MIE sim-
ulation results with an exact solution for two species when a set of 92 ODEs is
431
Table 12.1 Values of and for use in (12.69)
s
s,1

s,2

s,3

s,4

s,5

s,6

s
1 1 1
2 1 1 / 3 2 / 3
3 1 6 / 11 1 / 11 6 / 11
4 1 35 / 50 10 / 50 1 / 50 24 / 50
5 1 225 / 274 85 / 274 15 / 274 1 / 274 120 / 274
6 1 1624 / 1764 735 / 1764 175 / 1764 21 / 1764 1 / 1764 720 / 1764
solved simultaneously. Figure 12.3 shows how a change in the time step affects the
number of iterations and accuracy of the MIE solution in the 92-ODE case.
12.8 GEARS SOLUTION TO ODES
A more advanced method of solving stiff chemical ODEs is Gears method (Gear
1971). Gears method is accurate and elegant. A drawback of the original method
was its need to solve equations containing large matrices of partial derivatives.
This drawback prevented the use of the original code in three-dimensional models.
The application of sparse-matrix techniques improved the speed of Gears code
tremendously (e.g., Sherman and Hindmarsh 1980; Hindmarsh 1983; Jacobson
1994, 1995, 1998a; Jacobson and Turco 1994), but the use of sparse-matrix tech-
niques alone still did not permit the use of Gears code in three-dimensional models.
Only when sparse-matrix techniques were combined with computer optimization
techniques could Gears code be used to study problems in three dimensions (Jacob-
son 1994, 1995, 1998a; Jacobson and Turco 1994).
Gears method solves the backward differentiation formula (BDF). The BDF is
obtained by rst discretizing the time derivative of an ODE with
dN
i.t
dt

N
i.t
s.1
N
i.th
s.2
N
i.t2h

s.s
N
i.tsh
h
s
=
N
i.t
j =1
s. j
N
i.tj h
h
s
(12.69)
where s is the order of approximation of the method, and and are scalar mul-
tipliers, given in Table 12.1, that depend on the order of the method. Rearranging
(12.69) for an individual species and set of species gives
0 = N
i.t
+
s
j =1
s. j
N
i.tj h
+h
s
dN
i.t
dt
0 =
N
t
+
s
j =1
s. j

N
tj h
+h
s
d

N
t
dt
(12.70)
respectively. The solution to (12.70) is found by writing
P
t
(

N
t.m+1

N
t.m
) =
N
t.m
+
s
j =1
s. j

N
tj h
+h
s
d

N
t.m
dt
(12.71)
432
12.8 Gears solution to ODEs
where P
t
is a predictor matrix at time t, and d

N
t.m
dt is the rst derivative of

Nat
iteration m. Equation (12.71) can be rewritten as
P
t
L

N
t.m
=

B
t.m
(12.72)
where L

N
t.m
=

N
t.m+1

N
t.m
and
B
t.m
=
N
t.m
+
s
j =1
s. j

N
tj h
+h
s
d

N
t.m
dt
(12.73)
L

N
t.m
is solved from (12.72) by matrix decomposition and backsubstitution
during each iteration. Once it is solved, concentrations for the next iteration are
calculated with
N
t.m+1
=

N
t.m
+L

N
t.m
(12.74)
These values are substituted into (12.73) after each iteration, but only infrequently
into the predictor matrix. Upon iteration of (12.72)(12.74), L

N
t.m
approaches
zero. Convergence is checked with a local error test after each iteration and with
a global error test after the completion of all iterations of a time step. In the local
error test, a normalized root-mean-square error
NRMS
t.m
=
1
K
K
i =1
LN
i.t.m
R
tol
N
i.t.1
+ A
tol,t
2
(12.75)
is calculated, where LN
i.t.m
(molec. cm
3
) is the change in concentration of species
i at time step t during iteration m, K is the number of species, R
tol
is a constant
relative error tolerance, A
tol.t
is an absolute error tolerance for time step t (molec.
cm
3
), and N
i.t.1
is a concentration at the start of a time step. The relative error
tolerance controls errors relative to N
i.t.1
, and the absolute error tolerance controls
errors relative to xed concentrations. For pure relative-error-tolerance control,
A
tol,t
=0. For pure absolute-error-tolerance control, R
tol
=0. Typically, R
tol
=10
3
.
For gas chemistry, A
tol,t
10
3
10
7
molec. cm
3
. Proper selection or prediction of
A
tol,t
reduces the computer time in large models. Jacobson (1998a) gives a method
of predicting A
tol,t
.
If NRMS
t.m
is less than a specied parameter that varies with the order of
approximation, the local error test succeeds. If the local error test fails but NRMS
t.m
is decreasing relative to NRMS
t.m1
, iterations continue until the local error test
succeeds. If NRMS
t.m
increases upon iteration, convergence is not occurring, and
the matrix of partial derivatives is updated, and iterations continue. If the local
error test fails again, the time step is reduced until the test succeeds. Once the
local error test has succeeded, a global error test is performed to check whether a
433
cumulative normalized root-mean-square error, determined as
NRMS
t
=
1
K
K
i =1

m
LN
i.t.m
R
tol
N
i.t.1
+ A
tol,t
2
(12.76)
exceeds another parameter value that depends on the order of approximation. In
this equation,
m
LN
i.t.m
is the net change in species concentration during a time
step. If the global error check fails, a new time step is predicted at the same or one
order lower approximation. If the error check continues to fail, the time step is
reduced. If the global test succeeds, the time step was successful, and

N
t.m+1
-values
from the last iteration are set to nal concentrations. Every few successful time
steps, the time step and order of approximation are recalculated with a time-step
estimation scheme.
The predictor matrix on the left side of (12.72) expands to
P
t
I h
s
J
t
(12.77)
where I is the identity matrix, J
t
is a Jacobian matrix of partial derivatives, and
s
, which depends on the order of the method, was dened earlier. The Jacobian
matrix is
J
t
=

2

N
i.t.m
N
k.t.m
t
K.K
i.k=1
(12.78)
where K is the order of the matrix (number of species and ODEs). Matrices of
partial derivatives for chemical ODEs are sparse, and sparse-matrix techniques are
useful for treating them. Below, a matrix from a small equation set is shown, and
a technique of reducing matrix computations is discussed.
The reactions used for the example matrix are
NO+O
3

NO
2
+O
2
(k
1
) (12.79)
+ O
2
+MO
3
+M (k
2
) (12.80)
NO
2
+h
NO+
(J ) (12.81)
The corresponding rst derivatives of NO, NO
2
, O, and O
3
are
d[NO]
dt
= J [NO
2
] k
1
[NO][O
3
]
d[NO
2
]
dt
= k
1
[NO][O
3
] J [NO
2
]
d[O]
dt
= J [NO
2
] k
2
[O][O
2
][M]
d[O
3
]
dt
= k
2
[O][O
2
][M] k
1
[NO][O
3
]
(12.82)
The corresponding partial derivatives are as follows:
434
Partial derivatives of NO
2
[NO]
[NO]t
= k
1
[O
3
]

2
[NO]
[NO
2
]t
= J

2
[NO]
[O
3
]t
= k
1
[NO] (12.83)
Partial derivatives of NO
2
2
[NO
2
]
[NO]t
= k
1
[O
3
]

2
[NO
2
]
[NO
2
]t
= J

2
[NO
2
]
[O
3
]t
= k
1
[NO] (12.84)
Partial derivatives of O
2
[O]
[NO
2
]t
= J

2
[O]
[O]t
= k
2
[O
2
][M] (12.85)
Partial derivatives of O
3
2
[O
3
]
[NO]t
= k
1
[O
3
]

2
[O
3
]
[O]t
= k
2
[O
2
][M]

2
[O
3
]
[O
3
]t
= k
1
[NO] (12.86)
All other partial derivatives are zero.
For the set of four active species, the predictor matrix (P
t
) has the form
NO NO
2
O O
3
NO
NO
2
O
O
3
1 h
s
2
[NO]
[NO]t
h
s
2
[NO]
[NO
2
]t
h
s
2
[NO]
[O]t
h
s
2
[NO]
[O
3
]t
h
s
2
[NO
2
]
[NO]t
1 h
s
2
[NO
2
]
[NO
2
]t
h
s
2
[NO
2
]
[O]t
h
s
2
[NO
2
]
[O
3
]t
h
s
2
[O]
[NO]t
h
s
2
[O]
[NO
2
]t
1 h
s
2
[O]
[O]t
h
s
2
[O]
[O
3
]t
h
s
2
[O
3
]
[NO]t
h
s
2
[O
3
]
[NO
2
]t
h
s
2
[O
3
]
[O]t
1 h
s
2
[O
3
]
[O
3
]t
(12.87)
Substituting the partial derivatives from (12.83)(12.86) into (12.87) gives
NO NO
2
O O
3
NO
NO
2
O
O
3
1 h
s
(k
1
[O
3
]) h
s
(J ) 0 h
s
(k
1
[NO])
h
s
(k
1
[O
3
]) 1 h
s
(J ) 0 h
s
(k
1
[NO])
0 h
s
(J ) 1 h
s
(k
2
[O
2
][M]) 0
h
s
(k
1
[O
3
]) 0 h
s
(k
2
[O
2
][M]) 1 h
s
(k
1
[NO])
(12.88)
In (12.88), 5 out of the 16 matrix positions contain zeros. Thus, the initial
matrix ll-in is 1116. For larger chemical reaction sets, the order of the matrix is
larger, but the initial percentage ll-in is much smaller. A typical set of 90 species
and 200 chemical reactions has an initial ll-in of about 8 percent. The method
435
of implementing a sparse-matrix technique, described shortly, can be applied to a
matrix with any percentage of initial ll-in.
1 h
s
(k
1
[O
3
]) h
s
(J ) 0 h
s
(k
1
[NO])
h
s
(k
1
[O
3
]) 1 h
s
(J ) 0 h
s
(k
1
[NO])
0 h
s
(J ) 1 h
s
(k
2
[O
2
][M]) 0
h
s
(k
1
[O
3
]) 0 h
s
(k
2
[O
2
][M]) 1 h
s
(k
1
[NO])
L[NO]
t.m
L[NO
2
]
t.m
L[O]
t.m
L[O
3
]
t.m
B
NO.t.m
B
NO
2
.t.m
B
O.t.m
B
O
3
.t.m
(12.89)
which is solved either by Gaussian elimination or by matrix decomposition and
backsubstitution (e.g., Press et al. 1992). Because Gears method reuses the same
decomposed matrix for several backsubstitutions over the right side of (12.89),
decomposition plus backsubstitution is favored over Gaussian elimination.
During the decomposition of matrix equation (12.89), many multiplications
by zero occur. One sparse-matrix technique is to remove multiplications by zero
when such multiplications are known in advance. That is not enough. To reduce
computer time signicantly, the matrix should be reordered before multiplications
by zero are eliminated. An efcient way of reordering is to place species with the
most partial-derivative terms at the bottom and those with the fewest at the top of
the matrix (Jacobson 1994; Jacobson and Turco 1994). The use of this reordering
scheme requires that partial pivoting be avoided during decomposition. Such a
restriction is reasonable in that, if the decomposition should fail, the step size in
the numerical solver can be reduced. Reordering (12.89) gives
1 h
s
(k
2
[O
2
][M]) h
s
(J ) 0 0
0 1 h
s
(J ) h
s
(k
1
[O
3
]) h
s
(k
1
[NO])
0 h
s
(J ) 1 h
s
(k
1
[O
3
]) h
s
(k
1
[NO])
h
s
(k
2
[O
2
][M]) 0 h
s
(k
1
[O
3
]) 1 h
s
(k
1
[NO])
L[O]
t.m
L[NO
2
]
t.m
L[NO]
t.m
L[O
3
]
t.m
B
O.t.m
B
NO
2
.t.m
B
NO.t.m
B
O
3
.t.m
(12.90)
Reordering maximizes the number of multiplications by zero and minimizes
matrix ll-in during decomposition and backsubstitution. Table 12.2 shows that
sparse-matrix techniques can reduce the number of multiplications for a reaction
set of nearly 4000 reactions by about 99.995 percent, resulting in a computer
speedup of 15 000.
Although sparse-matrix techniques speed Gears solution method, computer
optimization techniques are also often needed to increase the speed of the chem-
istry solution in large models with 50500 chemical ODEs, 10
3
10
6
grid cells, and
integration times of several years. One optimization technique, useful on vector,
scalar, and parallel computers, is to divide the grid domain into blocks of 100500
436
Table 12.2 Reduction in array space and matrix operations due to reordering and removing
multiplications by zero for a chemistry set with 1427 gases and 3911 reactions under
daytime and nighttime conditions
After sparse reductions
Quantity Initial Day Percent reduction Night Percent reduction
Order of matrix 1 427 1 427 0 1 427 0
Initial ll-in 2 036 329 14 276 99.30 12 820 99.37
Final ll-in 2 036 329 17 130 99.16 14 947 99.27
Decomp. 1 967 595 901 47 596 99.995 36 974 99.996
Decomp. 2 1 017 451 9 294 99.09 7 393 99.27
Backsub. 1 1 017 451 9 294 99.09 7 393 99.27
Backsub. 2 1 017 451 6 409 99.37 6 127 99.40
The initial and nal ll-in are the initial and nal numbers of matrix positions lled. The last
four rows show the number of operations in each of four loops of matrix decomposition and
backsubstitution. The Initial column shows values before sparse-matrix reductions. The solver
was SMVGEAR II.
Source: Liang and Jacobson (2000).
grid cells each. For example, 100 000 cells can be divided into 200 blocks of 500
cells each. Chemical ODEs can be solved in blocks of 500 cells at a time instead of
serially, one at a time. When a grid is divided into blocks, every inner loop of the
computer program should be the grid-cell loop, with length 500.
The following example (in Fortran) demonstrates how array references are min-
imized when the inner loop is the grid-cell loop. In this example, NBIMOLEC is
the number (e.g., 150) of bimolecular reaction rates that are calculated, JSP1 is
the species number of the rst reactant, JSP2 is the species number of the second
reactant, KTLOOP is the number (e.g., 500) of grid cells in a grid block, RRATE is
the rate coefcient, CONC is a gas concentration, and TRATE is the reaction rate.
In nested loop A, where the inner loop is the grid-cell loop, only 150 references are
made to arrays JPROD1 and JPROD2.
Nested loop A
DO 105 NK = 1, NBIMOLEC
JSP1 = JPROD1(NK)
JSP2 = JPROD2(NK)
DO 100 K = 1, KTLOOP
TRATE(K,NK) = RRATE(K,NK)

CONC(K,JSP1)

CONC(K,JSP2)
100 CONTINUE
105 CONTINUE
Nested loop B
DO 105 K = 1, KTLOOP
DO 100 NK = 1, NBIMOLEC
JSP1 = JPROD1(NK)
JSP2 = JPROD2(NK)
TRATE(K,NK) = RRATE(K,NK)

CONC(K,JSP1)

CONC(K,JSP2)
100 CONTINUE
105 CONTINUE
437
0
50
100
150
200
10
14
10
12
10
10
10
8
10
6
10
4
0.01 1
N
o
.

o
f

b
a
c
k
s
u
b
s
t
i
t
u
t
i
o
n

c
a
l
l
s
Predictor value (s
2
)
Figure 12.4 Resulting number of backsubstitution calls
for different stiffness-predictor values from (12.91). The
more backsubstitution calls, the greater the stiffness of
the set of equations. The calculations were performed
with the computer code SMVGEAR II (Jacobson 1995).
In nested loop B, where the grid-cell loop is on the outside, the number of references
to JPROD1 and JPROD2 is 150 500. Thus, one pass through nested loop A is
much faster than one pass through nested loop B, regardless of the machine. On
vector machines, nested loop A also vectorizes better than does nested loop B.
When a grid domain is divided into blocks of grid cells, as described above,
another issue must be addressed. Chemistry in each grid cell of the block is iterated
in the same manner as in each other grid cell in the block. As such, iterations must
continue in all grid cells in the block until the grid cell with stiffest chemistry con-
verges. This results in many unnecessary iterations in grid cells where convergence
occurs faster than in other grid cells. Excess iterations can be minimized by reorder-
ing grid cells among all grid blocks each time interval, according to stiffness. For
example, if 20 000 cells are solved 500 at a time, the 500 stiffest cells can be placed
together in a block, the next 500 stiffest can be placed together, and so on. This
way, cells that require the most iterations are grouped together.
To reorder cells according to stiffness each time interval, a stiffness predictor is
needed. One such predictor is
S
p
=
1
K
K
i =1
dN
i.t
dt
N
i.t
+ A
tol,t
2
(12.91)
(Jacobson 1995), where S
p
is proportional to stiffness, K is the number of ODEs
(and species), A
tol,t
is the absolute error tolerance at time t, N
i,t
is the species
concentration, and dN
i,t
dt is the rst derivative of N
i,t
. The larger the value of S
p
,
the stiffer the equations in the grid cell. Figure 12.4 shows how (12.91) predicted
the stiffness in an example simulation.
In sum, Gears method is accurate for solving atmospheric chemistry problems.
When the method is combined with sparse-matrix and computer optimization
438
12.9 Family solution to ODEs
techniques, the equations can be solved over large, three-dimensional model
domains for long integration periods.
12.9 FAMILY SOLUTION TO ODES
The last numerical method of solving chemical ODEs discussed is the family method
(e.g., Crutzen 1971; Turco and Whitten 1974; Austin 1991; Elliott et al. 1993;
Jacobson 1994). This method is less accurate than Gears method for the same
time step and must be tuned for each application. It is fast and useful for several
applications.
The theory behind the family method is that some groups of gases, or families,
exist in which atoms transfer quickly among species in the family but are lost only
slowly from the family. Example families are the odd oxygen, odd hydrogen, odd
nitrogen, and odd chlorine families. Some of the major species in these families are
Odd oxygen: [O
T
] = [O] +[O(
1
D)] +[O
3
] +[NO
2
] (12.92)
Odd hydrogen: [HO
T
] = [OH] +[HO
2
] +[H
2
O
2
] (12.93)
Odd nitrogen: [NO
T
] = [NO] +[NO
2
] +[NO
3
] (12.94)
Odd chlorine [Cl
T
] = [Cl] +[ClO] +[ClO
2
] (12.95)
For example, oxygen atoms in the odd-oxygen family cycle among the species
atomic oxygen, excited atomic oxygen, and ozone by the reactions,
O
3
+h O
2
+
O(
1
D)
O
2
+h
+

O
(12.96)
O(
1
D)
M
+O
2
+MO
3
+M
Cycling of oxygen atoms among these reactions is fast, whereas loss of oxygen
atoms out of this group of reactions is slower.
The family solution to ODEs for a time step requires four steps. First, rates of
production and loss of individual species are calculated frominitial concentrations.
These rates are then summed across a family. The family concentration is then
advanced with a forward Euler approximation applied to the summed production
and loss terms. Finally, species concentrations are repartitioned in the family for
the next time step.
Suppose a system contains four species (A, B, C, and D) and two families,
[Fam
1
] = [A] +[B] +[C] [Fam
2
] = [D] (12.97)
If the reactions
A B (k
a
) B C (k
b
) C D (k
c
) (12.98)
439
occur in the system, the rst step in the family solution method is to calculate rates
of production and loss of the individual species, A, B, C, and D:
d[A]
dt
= k
a
[A]
d[B]
dt
= k
a
[A] k
b
[B]
d[C]
dt
= k
b
[B] k
c
[C]
d[D]
dt
= k
c
[C]
(12.99)
The second step is to sum rates of production and loss across each family. Taking
the time derivative of (12.97) and substituting (12.99) into the result gives
d[Fam
1
]
dt
=
d[A]
dt
+
d[B]
dt
+
d[C]
dt
= k
c
[C]
d[Fam
2
]
dt
=
d[D]
dt
= k
c
[C] (12.100)
The third step is to advance the family concentration over a time step with a forward
Euler. For example,
[Fam]
t
= [Fam]
th
+h
d[Fam]
th
dt
(12.101)
[Fam
1
]
t
= [Fam
1
]
th
hk
c
[C]
th
[Fam
2
]
t
= [Fam
2
]
th
+hk
c
[C]
th
(12.102)
Fourth, species concentrations must be extracted from family concentrations. One
method of extracting species concentrations is to calculate and apply partitioning
ratios. Such ratios are calculated by summing the concentrations of the species in
the family as
[Fam
1
]
t
= [A]
t
+[B]
t
+[C]
t
= [A]
t
1 +
[B]
t
[A]
t
+
[C]
t
[A]
t
(12.103)
Individual species concentrations are then found from the family concentrations
with
[A]
t
=
[Fam
1
]
t
1 +
[B]
t
[A]
t
+
[C]
t
[A]
t
[B]
t
= [A]
t
[B]
t
[A]
t
[C]
t
= [A]
t
[C]
t
[A]
t
(12.104)
In these equations, the partitioning ratios [B]
t
[A]
t
and [C]
t
[A]
t
are currently
unknown. In the case of species D, its nal concentration is that of the family it
resides in, since no other species is in the family. Thus, [D]
t
= [Fam
2
]
t
= hk
c
[C]
th
.
The simplest but least accurate way to estimate partitioning ratios is to assume
that each species in the family is in steady state. With this method, the individual
rates of reaction of species B and C are set to zero:
d[B]
dt
= k
a
[A] k
b
[B] = 0
d[C]
dt
= k
b
[B] k
c
[C] = 0 (12.105)
The partitioning ratios of B and C are then estimated as
[B]
t
[A]
t
[B]
[A]
=
k
a
k
b
[C]
t
[A]
t
[C]
[A]
=
[C]
[B]
[B]
[A]
=
k
b
k
c
k
a
k
b
=
k
a
k
c
(12.106)
440
12.9 Family solution to ODEs
respectively. Substituting (12.106) into (12.104) gives the nal species concentra-
tions of A, B, and C, respectively, as
[A]
t
=
[Fam
1
]
t
1 +
k
a
k
b
+
k
a
k
c
[B]
t
=
[Fam
1
]
t
k
a
k
b
1 +
k
a
k
b
+
k
a
k
c
[C]
t
=
[Fam
1
]
t
k
a
k
c
1 +
k
a
k
b
+
k
a
k
c
(12.107)
A second way to estimate partitioning ratios is to linearize the ODEs of the
individual species as
d[A]
t
dt
= k
a
[A]
t
d[B]
t
dt
k
a
[A]
th
k
b
[B]
t
d[C]
t
dt
k
b
[B]
th
k
c
[C]
t
(12.108)
and then to integrate each equation. The results are
[A]
t
= [A]
th
e
k
a
h
[B]
t
[B]
th
e
k
b
h
+
k
a
[A]
th
k
b
1 e
k
b
h
[C]
t
[C]
th
e
k
c
h
+
k
b
[B]
th
k
c
1 e
k
c
h
(12.109)
These estimates are substituted into the right sides of (12.104).
A third way to estimate partitioning ratios is to nite-difference the ODEs as
[A]
t
[A]
th
h
= k
a
[A]
t
[B]
t
[B]
th
h
= k
a
[A]
t
k
b
[B]
t
[C]
t
[C]
th
h
= k
b
[B]
t
k
c
[C]
t
(12.110)
and then to rearrange the equations into a matrix equation as
1+hk
a
0 0
hk
a
1+hk
b
0
0 hk
b
1+hk
c
[A]
t
[B]
t
[C]
t
[A]
th
[B]
th
[C]
th
(12.111)
Solving the matrix equation for [A]
t
, [B]
t
, and [C]
t
then substituting the results into
the right sides of (12.104) gives nal partitioned concentrations.
In the example above, the reactions were unimolecular. Most chemical reactions
are bimolecular. When bimolecular reactions are used, the rst derivatives need to
be linearized for use in (12.111). If the bimolecular reaction A +C B +D (k
ac
)
is added to (12.98), the linearized nite-difference forms for A, B, and C are
[A]
t
[A]
th
h
= k
a
[A]
t
k
ac
[A]
t
[C]
th
(12.112)
[B]
t
[B]
th
h
= k
a
[A]
t
k
b
[B]
t
+0.5k
ac
([A]
t
[C]
th
+[A]
th
[C]
t
) (12.113)
[C]
t
[C]
th
h
= k
b
[B]
t
k
c
[C]
t
k
ac
[A]
th
[C]
t
(12.114)
441
respectively. The resulting matrix is (Jacobson 1994)
1 +hk
a
+hk
ac
[C]
th
0 0
hk
a
0.5(hk
ac
[C]
th
) 1 +hk
b
0.5(hk
ac
[A]
th
)
0 hk
b
1 +hk
c
+hk
ac
[A]
th
[A]
t
[B]
t
[C]
t
[A]
th
[B]
th
[C]
th
(12.115)
Solving this matrix equation gives concentrations that can be used to estimate
partitioning ratios.
The advantages of the family method are that it is fast, since it can use a long
time step, and it may be accurate for moderate- to low-stiffness systems. The disad-
vantages are that the families need to be designed carefully and validated for each
set of chemistry, and the accuracy of the method decreases with increasing stiffness.
Also, although families are generally long-lived, the forward Euler solution for the
family often results in negative concentrations. In such cases, the time step must be
reduced and the family concentration must be recalculated. This reduces the speed
advantage of the family method.
12.10 SUMMARY
In this chapter, methods of solving chemical rst-order ordinary differential equa-
tions were described. These included analytical, explicit Taylor series, forward
Euler, backward Euler, simple exponential, quasi-steady-state, multistep implicit
explicit, Gear, and family methods. A good solution scheme is stable, accurate,
mass-conserving, positive denite, and fast. One code that ts these qualities is
Gears code combined with sparse-matrix and vectorization techniques. Stiff ODE
codes can be used to solve aqueous as well as gas-phase chemical equations. Such
equations are discussed in Chapter 19.
12.11 PROBLEMS
12.1 Derive ananalytical solutionfor the loss of molecular oxygenvia the reaction
O
2
+h O+O. If oxygen is destroyed by this reaction but not re-created,
how long will it take for its concentration to decrease to 10 percent of its
initial value? Use a constant stratospheric photolysis rate coefcient from
Appendix Table B.4.
12.2 Given reactions A+h B +C (J
1
) and B +h C +C (J
2
), where J
1
and
J
2
are photolysis coefcients and where [A], [B], and [C] are concentrations
of A, B, and C:
(a) Write out the chemical rate equations for A, B, and C (e.g., d[A]dt . . .)
(b) Find the time-dependent analytical solutions for each species from
the initial values [A]
th
, [B]
th
, and [C]
th
. Assume J
1
and J
2
are constant
in time.
(c) As time approaches innity (h ), what is [C]? What is [C]
h
when
J
1
= 0?
442
Table 12.3 Reactions and rate coefcients for
Problem 12.3
Reaction Rate-coefcient expression
O
O
2
.M
O
3
k
1
= 1.4 10
3
e
1175T
NO+O
3
NO
2
+O
2
k
2
= 1.8 10
12
e
1370T
NO
2
+h NO+O J = 1.7 10
2
In the rst reaction, only O is included in the rate-
coefcient expression, since the rate coefcient has
already been multiplied by O
2
and M. T is absolute tem-
perature (K).
0
210
9
410
9
610
9
810
9
110
10
1.210
10
110
12
1.00110
12
1.00210
12
1.00310
12
1.00410
12
1.00510
12
0 400 800 1200 1600 2000
O
3
,

O
(
3
P
)
,

N
O
2

(
m
o
l
e
c
u
l
e
s

c
m
3
)
N
O

(
m
o
l
e
c
u
l
e
s

c
m
3
)
Time (s)
NO
O
3
O(
3
P) 10
6
NO
2
Figure 12.5 Exact and modeled solution to Problem 12.3. Exact concentrations
are represented by continuous lines. Solutions from the MIE method are repre-
sented by symbols.
12.3 Write a program to solve chemical ODEs with the MIE algorithm. Test the
program with the three reactions shown in Table 12.3, assuming that T =
298 K. Assume that the initial concentrations of O
3
, O(
3
P), NO, and NO
2
are 0, 0, 10
12
, and 10
10
molec. cm
3
, respectively. Do not solve for O
2
. Use
a time step of h = 10 s. The results should exactly match those shown in
Fig. 12.5.
12.4 Write a program to solve chemical ODEs with the simple exponential
method. Test the program with the three reactions in Table 12.3, assum-
ing T =298 K. Assume that initial concentrations of O
3
, O(
3
P), NO, and NO
2
are 0, 0, 10
12
, and 10
10
molec. cm
3
, respectively. Find a time step size, if
any, that gives the results shown in Fig. 12.5. Discuss what happens to the
solution at increasingly larger time steps.
12.5 Repeat Problem 12.4 with a forward Euler approximation. Determine
the time step size at which the solution becomes unstable (concentrations
443
Table 12.4 Typical volume-mixing ratios of several gases at two altitudes
Volume mixing ratio Volume mixing ratio
(fraction) (fraction)
Gas 25 km 0 km Gas 25 km 0 km
N
2
0.7808 0.7808 CH
3
COOH 3.0 (9)
O
2
0.2095 0.2095 CH
3
COCH
3
7.5 (10)
O
3
8.0 (6) 4.0 (8) C
5
H
8
(ISOP) 2.0 (10)
H
2
6.0 (7) 6.0 (7) CH
3
Cl 3.6 (10) 6.0 (10)
H
2
O 3.1 (6) 0.01 CH
3
CCl
3
2.0 (11) 1.4 (10)
H
2
O
2
2.1 (10) 3.0 (9) CCl
4
1.3 (11) 1.1 (10)
NO 1.1 (9) 5.0 (12) CFCl
3
5.0 (11) 2.7 (10)
NO
2
1.9 (9) 4.0 (11) CF
2
Cl
2
2.2 (10) 4.8 (10)
N
2
O 1.5 (7) 3.0 (7) CF
2
ClH 7.0 (11) 1.2 (10)
HNO
3
6.0 (9) 5.0 (10) CFCl
2
CF
2
Cl 3.0 (11) 7.0 (11)
HO
2
NO
2
3.4 (10) 1.0 (11) HCl 1.3 (9) 9.0 (11)
CO 4.0 (8) 1.1 (7) ClONO
2
8.0 (10)
CO
2
3.75 (4) 3.6 (4) HOCl 3.3 (11)
CH
4
1.0 (6) 1.8 (6) CH
3
Br 1.0 (11) 1.2 (11)
C
2
H
6
6.0 (10) BrONO
2
2.0 (12)
C
3
H
8
4.8 (11) HBr 3.0 (12)
C
2
H
4
2.1 (11) SO
2
1.0 (11) 5.0 (11)
C
3
H
6
6.0 (12) CH
3
SCH
3
1.0 (10)
HCHO 2.0 (10) H
2
S 5.0 (11)
CH
3
CHO 1.6 (10) OCS 1.0 (10) 5.0 (10)
HCOOH 1.8 (9) CS
2
2.0 (11) 1.0 (10)
CH
3
OOH 1.2 (9)
6.0 (6) means 6.0 10
6
. Multiply the volume mixing ratio of a gas (molecules of gas
per molecule of dry air) by the number concentration of dry air (molec. cm
3
) to obtain the
number concentration of the gas. The 0-km mixing ratios correspond to clean tropospheric
conditions.
oscillate between positive and negative values or explode to large positive
or negative numbers).
12.6 Repeat Problem 12.4 with a backward Euler approximation. Discuss differ-
ences between the results from Problems 12.5 and 12.6.
12.13 MODELING PROJECT
Using the program developed for Problem 12.3, read in the equations and rate
coefcients from the computer le developed for Problems 11.4 and 10.7. Run a
four-day simulation, starting at 5:00 a.m. on the rst day. Calculate initial num-
ber concentrations for necessary species from volume mixing ratios shown in
Table 12.4. If a stratospheric mechanism was chosen for Problem 11.4, assume
T = 192 K and p
d
= 25 hPa, and use the mixing ratios at 25 km in Table 12.4. If
a near-surface mechanism was chosen, assume T = 298.15 K and p
d
= 1013 hPa,
444
and use the mixing ratios at 0 km, which correspond to clean lower-tropospheric
conditions. Assume all initial mixing ratios not given in the table are zero. Once
a baseline simulation has been run, vary temperature, pressure, and initial mixing
ratios to test the sensitivity of model predictions to these parameters. Display time-
series plots of species mixing ratios from the baseline and sensitivity test cases on
the same graphs. Discuss results and their signicance.
445
13
Particle components, size distributions,
and size structures
P
arti cles in the atmosphere vary in size and composition. In this chapter,
their importance and treatment in numerical models are discussed. The vari-
ation of particle concentration with size can be simulated with a continuous or
discrete size distribution. Three continuous distributions discussed here are the
lognormal, MarshallPalmer, and modied gamma distributions. These distribu-
tions can be discretized over a model size grid, which consists of size bin increments
in radius space. The variation of a size distribution with time is simulated with a size
structure. Five structures discussed here are the full-stationary, full-moving, qua-
sistationary, hybrid, and moving-center structures. In later chapters, the treatment
of time-dependent processes that affect particle size distributions and composition
are examined. These include emission, nucleation, coagulation, growth, evapora-
tion, chemical equilibrium, interaction with clouds, aqueous chemistry, rainout,
washout, sedimentation, and dry deposition.
13.1 INTRODUCTION TO PARTICLES
An aerosol is an ensemble of solid, liquid, or mixed-phase particles suspended in
air. An aerosol particle is a single such particle. A hydrometeor is an ensemble of
liquid, solid, or mixed-phase particles containing primarily water, suspended in or
falling through the air. A hydrometeor particle is a single such particle. Examples
of hydrometeor particles are fog drops, cloud drops, drizzle, raindrops, ice crys-
tals, graupel, snowakes, and hailstones. The main difference between an aerosol
particle and a hydrometeor particle is that the latter contains much more water
than does the former. Whereas almost all mass within a cloud drop and raindrop is
liquid water, aerosol particles contain large fractions of other material as well. The
sizes and number concentrations of aerosol particles and hydrometeor particles
also differ. Table 13.1 compares characteristics of gas molecules, aerosol particles,
and hydrometeor particles.
Aerosol particles are important because they affect health, air quality, cloud for-
mation, meteorology, and climate. Submicrometer aerosol particles (those smaller
than 1 m in diameter) affect human health by penetrating to the deepest part of
human lungs. Aerosol particles 0.21 m in diameter that contain sulfate, nitrate,
and organic carbon scatter light efciently. Aerosol particles smaller than 1 mthat
contain black carbon absorb light efciently. Aerosol absorption and scattering
446
13.2 Aerosol, fog, and cloud composition
Table 13.1 Characteristics of gases, aerosol particles, and hydrometeor particles
Typical diameter Number concentration Mass concentration
(m) (cm
3
) (g m
3
)
Gas molecules 0.0005 2.45 10
19
1.2 10
9
Aerosol particles
Small -0.2 10
3
10
6
-1
Medium 0.21.0 110
4
-250
Large 1.0100 -110 -500
Hydrometeor particles
Fog drops 1020 1500 10
4
510
5
Average cloud drops 10200 -101000 -10
5
510
6
Large cloud drops 200 -110 -10
5
510
6
Drizzle 400 0.1 10
5
510
6
Small raindrops 1000 0.01 10
5
510
6
Medium raindrops 2000 0.001 10
5
510
6
Large raindrops 8000 -0.001 10
5
510
6
Data are for typical lower-tropospheric conditions.
affect (1) radiative energy uxes, which affect temperatures, and (2) photolysis,
which affects the composition of the atmosphere. Aerosol particles also serve as
sites on which cloud drops form. In fact, without aerosol particles, clouds would
rarely form in the atmosphere. Finally, aerosol particles serve as sites on which
chemical reactions take place and as sites for trace gases to condense upon or
dissolve within.
13.2 AEROSOL, FOG, AND CLOUD COMPOSITION
Aerosol particles enter the atmosphere in one of only two ways: emission and homo-
geneous nucleation (Chapter 14). New particles emitted into the atmosphere are
primary particles. New particles formed by homogeneous nucleation of gases are
secondary particles. Homogeneous nucleation is a gas-to-particle conversion pro-
cess because it results in the change of state of a gas to a liquid or solid. Other gas-
to-particle conversion processes include heterogeneous nucleation (Chapter 14),
condensation (Chapter 16), dissolution (Chapter 17), and heterogeneous reaction
(Chapter 17). These processes, though, do not result in the formation of new par-
ticles. They result in the addition of mass and new chemicals to existing particles.
When new particles are emitted or formed by homogeneous nucleation, they are
externally mixed. Externally mixed particles are those that contain only the com-
ponents they are emitted with or homogeneously nucleated with. As particles age,
coagulation (the collision and coalescence of particles) and gas-to-particle conver-
sion internally mix particles. Internally mixed particles are particles that contain
chemicals in addition to those the particles originally contained from emission or
homogeneous nucleation.
447
Particle components, size distributions, and size structures
Table 13.2 Liquids, ions, and solids found in atmospheric aerosol particles
Liquids
Sulfuric acid H
2
SO
4
(aq) Formic acid HCOOH(aq)
Nitric acid HNO
3
(aq) Acetic acid CH
3
COOH(aq)
Hydrochloric acid HCl(aq) Ammonia NH
3
(aq)
Carbonic acid H
2
CO
3
(aq) Hydrogen peroxide H
2
O
2
(aq)
Sulfurous acid H
2
SO
3
(aq) Organic matter C, H, O, N(aq)
Ions
Bisulfate ion HSO
4
Sulte ion SO
3
2
Sulfate ion SO
4
2
Hydrogen ion H
+
Nitrate ion NO
3
Ammonium ion NH
4
+
Chloride ion Cl
Sodium ion Na
+
Bicarbonate ion HCO
3
Calcium ion Ca
2+
Carbonate ion CO
3
2
Magnesium ion Mg
2+
Bisulte ion HSO
3
Potassium ion K
+
Solids
Ammonium bisulfate NH
4
HSO
4
(s) Potassium sulfate K
2
SO
4
(s)
Ammonium sulfate (NH
4
)
2
SO
4
(s) Potassium nitrate KNO
3
(s)
Triammonium bisulfate (NH
4
)
3
H(SO
4
)
2
(s) Potassium chloride KCl(s)
Ammonium nitrate NH
4
NO
3
(s) Potassium bicarbonate KHCO
3
(s)
Ammonium chloride NH
4
Cl(s) Potassium carbonate K
2
CO
3
(s)
Sodium bisulfate NaHSO
4
(s) Black carbon BC(s)
Sodium sulfate Na
2
SO
4
(s) Organic matter C, H, O, N(s)
Sodium nitrate NaNO
3
(s) Silicon dioxide SiO
2
(s)
Sodium chloride NaCl(s) Iron (III) oxide Fe
2
O
3
(s)
Calcium sulfate CaSO
4
(s) Aluminum oxide Al
2
O
3
(s)
Calcium nitrate Ca(NO
3
)
2
(s) Lead suboxide Pb
2
O(s)
Calcium chloride CaCl
2
(s) Tire particles
Calcium carbonate CaCO
3
(s) Pollen
Magnesium sulfate MgSO
4
(s) Spores
Magnesium nitrate Mg(NO
3
)
2
(s) Bacteria
Magnesium chloride MgCl
2
(s) Viruses
Magnesium carbonate MgCO
3
(s) Plant debris
Potassium bisulfate KHSO
4
(s) Meteoric debris
Aerosol particles can contain chemicals in the liquid and/or solid phases and
molecules that are inorganic and/or organic. When the relative humidity is high,
the most abundant chemical in aerosol particles is typically liquid water. A solvent
is a liquid in which gases dissolve. Because of its quantity within aerosol particles
and the ability of many gases to dissolve in it, liquid water is generally considered
the main solvent in aerosol particles. A substance that dissolves in a solvent is
a solute. Solute and solvent, together, make up a solution. Dissolved substances
can dissociate or remain undissociated in solution. A dissociated species has a
positive (cationic) or negative (anionic) electric charge. An undissociated species
has a neutral charge. Suspended materials, such as solids, may be mixed in a solution
but are not part of the solution.
Table 13.2 lists several undissociated liquids, ions, and solids found in atmo-
spheric aerosol particles. Appendix Table B.5 gives a more complete list that also
448
13.3 Discrete size distributions
includes components found in fog and clouds and identies gas-phase precursors.
The sources of aqueous sulfuric acid are homogeneous nucleation (Chapter 14),
condensation (Chapter 16), aqueous reaction (Chapter 19), and some direct emis-
sion (Chapter 14). The sources of aqueous nitric acid are primarily dissolution
of gas-phase nitric acid (Chapter 17) and some emission. The sources of aque-
ous hydrochloric acid are emission and dissolution of gas-phase hydrochloric acid.
Carbonic acid originates mostly from dissolution of carbon dioxide gas into water
within aerosol particles. Sulfurous acid originates primarily from dissolution of
sulfur dioxide gas. Formic acid, acetic acid, ammonia, and hydrogen peroxide sim-
ilarly originate from dissolution of their gas-phase counterparts. Liquid organic
matter originates from emission, condensation, and dissolution.
Ions within aerosol particles originate from dissociation of liquids or solids.
Many solids (e.g., ammonium sulfate, potassium nitrate) form from chemical reac-
tions among ions in solution. Some solids (e.g., calcium carbonate, silicon dioxide,
some iron (III) oxide) are emitted naturally in wind-blown soil dust. Other solids
(e.g., pollen, spores, bacteria, viruses, plant debris) are biological matter lifted by
the wind. Still others (e.g., black carbon, lead suboxide, some iron (III) oxide) are
emitted anthropogenically.
13.3 DISCRETE SIZE DISTRIBUTIONS
The rst step in designing algorithms to simulate particle size and composition in
an atmospheric model is to choose an initial particle size distribution. A particle
size distribution gives the number concentration, surface-area concentration, vol-
ume concentration, or mass concentration of particles as a function of radius or
diameter. Particle radii are usually expressed in micrometers (microns) for aerosol
particles and cloud drops, and millimeters for raindrops.
A size distribution may be continuous or discrete (e.g., Friedlander 1977). A
continuous distribution is one in which the variation of particle concentration
with radius is represented by a continuous function. A discrete distribution is one
in which concentrations are distributed over increments in radius space, called size
bins (or sections).
Some atmospheric models assume the aerosol size distribution is continuous
(e.g., Hulburt and Katz 1964; McGraw and Saunders 1984; Pratsinis 1988;
Binkowski and Shankar 1995; Terry et al. 2001). Others assume the aerosol size
distribution is discrete (e.g., Turco et al. 1979; Gelbard and Seinfeld 1980; Toon
et al. 1988; Jacobson 1997a,b; Meng et al. 1998; Zhang et al. 2004). Zhang et al.
(1999) compare some algorithms developed under these different assumptions.
The main advantage of a continuous distribution is that it reduces computer time
because relatively few parameters are stored and calculated relative to a discrete
distribution. The accuracy of aerosol dynamics with a continuous distribution can
also be high for many idealized cases because analytical solutions are available
in such cases. However, when aerosol particles contain multiple components and
several physical and chemical processes are treated simultaneously, the relatively
449
sparse information stored with continuous-distribution techniques increases the
error of such techniques relative to discrete-distribution techniques. For complex
aerosol interactions where computer time is not a major constraint, the use of a
discrete distribution with high resolution is ideal for treating aerosol dynamics.
A size bin in a discrete distribution has a lower and an upper edge radius (or
diameter). All particles with radii between the lower and upper edges of a size bin
are assumed to reside in the bin and to have the same diameter as each other. In
reality, each particle in the atmosphere has a distinct diameter. The lower atmo-
sphere contains on the order of 10
3
10
6
particles per cubic centimeter of air, and
particle diameters vary by over six orders of magnitude, from 0.001 m (nucleated
aerosol particles) to 5000 m (large dust particles). Each particle can contain from
1 to 1000 components.
Example 13.1
An idealized particle size distribution might consist of 10000 particles of
radius between 0.005 and 0.5 m, 100 particles of radius between 0.5 and 5.0
m, and 10 particles of radius between 5.0 and 50 m. The particles can be
naturally distributed into a model that uses three size bins.
The number of particle size bins and components in a model is limited for two
reasons. First, a typical three-dimensional model contains 10
3
10
6
grid cells, and
the inclusion of too many particle sizes and/or components results in computer-time
requirements that exceed those available. Second, the minimum storage require-
ment for particle concentrations over a model grid domain is one array with a
dimension equal to the number of grid cells in the domain multiplied by the num-
ber of size bins and the number of components. This dimension can quickly surpass
the central memory capabilities of many computers.
Example 13.2
With 10
5
grid cells, 100 size bins, and 100 components per size bin, the central
memory requirement, just to store one array for concentration, is one gigaword
(10
9
words). This translates to 8 gigabytes when 1 word = 8 bytes.
Limiting the number of sizes (or size bins) in a model is often the easiest method
of reducing computer-time and memory limits. Reducing the number of size bins
without losing resolution is difcult because particle sizes vary by many orders
of magnitude. To cover all particle sizes with a xed number of bins, the bins
should be spread geometrically over the size range of interest. A discrete geomet-
ric distribution, called the volume-ratio size distribution, is discussed below. With
450
v
1
V
rat
v
1
V
rat
v
2
V
rat
v
i1
i = 1 i = 2 i = 3 i
Figure 13.1 Variation in particle size with the
volume-ratio size distribution. Below each
particle is an expression for its single-particle
volume.
this distribution, the volume of particles in a size bin equals the volume of par-
ticles in the next smallest size bin multiplied by a constant volume ratio, V
rat
.
Thus,
i
= V
rat
i 1
for i = 1. . . . . N
B
(13.1)
where
i
is the volume (cm
3
) of a single particle in a size bin i,
i1
is the vol-
ume of a single particle in the next smaller size bin, and N
B
is the number of
size bins (e.g., Toon et al. 1988). The volume in a size bin can also be expressed
as
i
=
1
V
i 1
rat
for i = 1. . . . . N
B
(13.2)
where
1
is the volume of a single particle in the smallest size bin. Figure 13.1
illustrates how particle sizes are distributed in the volume-ratio distribution.
For many model applications, it is convenient to consider particles as spherical.
In such cases, the average volume of a single particle in any bin i is related to its
average diameter d
i
by
i
= d
3
i
6. In reality, particles have a variety of shapes.
Soot particles are porous and amorphous, whereas ice particles are crystalline. A
highly resolved model can consider several particle types, each with a different size
distribution, composition distribution, and shape distribution. However, particles
are often (but not always) assumed to be spherical, since that may be closer to an
average particle shape in the atmosphere than any other shape. Spherical geometry
is also easier to treat numerically than other geometries.
In (13.1) and (13.2), V
rat
must be specied in advance. A disadvantage of the
equations is that if V
rat
is specied incorrectly for a given N
B
, the resulting volume
of particles in the largest size bin,
N
B
may be smaller or larger than desired. To
avoid this problem, V
rat
can be precalculated from a specied
N
B
with
V
rat
=
N
B
1(N
B
1)
=
d
N
B
d
1
3(N
B
1)
(13.3)
451
where d
1
is the diameter of particles in the smallest bin, and d
N
B
is the diameter
of particles in the largest bin, assuming the particles are spherical. If, instead,
N
B
,
d
N
B
, and V
rat
are specied, but N
B
is not, then N
B
can be found with
N
B
= 1 +
ln
(d
N
B
d
1
)
3
ln V
rat
(13.4)
Example 13.3
Equation (13.3) predicts that, for 30 size bins to cover a preset diameter range
from 0.01 to 1000 m, we have V
rat
= 3.29.
Example 13.4
If d
1
= 0.01 m, d
N
B
= 1000 m, and V
rat
= 4, (13.4) implies that 26 size bins
are needed. If V
rat
=2, then 51 size bins are needed to cover the same diameter
range.
The volume width of a size bin in the volume-ratio distribution is derived by
rst assuming that the average single-particle volume in a bin is
i
=
1
2
(
i.hi
+
i.lo
) (13.5)
where
i.hi
and
i.lo
are the volumes of the largest and smallest particles in the bin,
respectively. The volume of the largest particles in the bin is limited by
i.hi
= V
rat
i.lo
(13.6)
i.lo
=
2
i
1 + V
rat
(13.7)
The volume width of a size bin is
L
i
=
i.hi
i.lo
=
2
i +1
1 + V
rat
2
i
1 + V
rat
=
2
i
(V
rat
1)
1 + V
rat
(13.8)
and the corresponding diameter width is
Ld
i
= d
i.hi
d
i.lo
=
13
13
i.hi

13
i.lo
= d
i
2
13
V
13
rat
1
(1 + V
rat
)
13
(13.9)
The average volume of one particle in a size bin is one parameter used in a
discrete size distribution. Four other parameters are the number, volume, area, and
452
mass concentrations. The number concentration n
i
(particles cm
3
) is the num-
ber of particles in a size bin per unit volume of air, the volume concentration v
i
(cm
3
cm
3
) is the volume of particles in a bin per unit volume of air, the surface-area
concentration a
i
(cm
2
cm
3
) is the surface area of particles in a bin per unit volume
of air, and the mass concentration m
i
(g m
3
) is the mass of particles in a bin per
unit volume of air. Number concentration is related to volume concentration and
single-particle volume in a bin by
n
i
=
v
i
i
(13.10)
The total number concentration of particles, summed over all size bins in a distri-
bution, is
N
D
=
N
B
i =1
n
i
(13.11)
Individual particles usually consist of a mixture of a few to hundreds of compo-
nents. For modeling purposes, the relationship between total volume concentration
and component volume concentrations v
q.i
(cm
3
cm
3
) is
v
i
=
N
V
q=1
v
q.i
(13.12)
where N
V
is the number of components within each particle.
Assuming that particles are spherical, the surface-area concentration (cm
2
cm
3
)
of particles in a size bin is
a
i
= n
i
4r
2
i
= n
i
d
2
i
(13.13)
where r
i
= d
i
2 is particle radius. The surface-area concentration is used for
estimating particle surface area available for gas condensation or heterogeneous
reaction.
The mass concentration of particles in a size bin is
m
i
=
N
V
q=1
m
q.i
= c
m
N
V
q=1
q
v
q.i
= c
m
p.i
N
V
q=1
v
q.i
= c
m
p.i
v
i
(13.14)
where m
q.i
= c
m
q
v
q.i
is the mass concentration (micrograms per cubic meter of
air) of component q in particles of size i, c
m
(= 10
12
) converts g cm
3
to g m
3
,
q
is the mass density of q (grams per cubic centimeter of component), and
p.i
is
the volume-averaged mass density of a particle in bin i (grams per cubic centimeter
of particle). The volume-averaged density changes continuously and is found by
453
equating the third and fourth terms in (13.14). The result is
p.i
=
N
V
q=1
(
q
v
q.i
)
N
V
q=1
v
q.i
(13.15)
Example 13.5
If the mass concentrations of water and sulfate are m
q.i
= 3.0 and 2.0 g
m
3
, respectively, in particles with an average diameter of d
i
= 0.5 m,
calculate the volume concentration of each component and the mass, vol-
ume, area, and number concentrations of whole particles in the size bin.
Assume the densities of water and sulfate are
q
= 1.0 and 1.83 g cm
3
,
respectively.
SOLUTION
Fromm
q.i
= c
m
:
q.i
q
, the volume concentrations of water and sulfate are :
q.i
=
3 10
12
and 1.09 10
12
cm
3
cm
3
, respectively. From (13.14), m
i
= 5.0 g
m
3
. From (13.12), :
i
= 4.09 10
12
cm
3
cm
3
. From
i
= d
3
i
6,
i
= 6.54
10
14
cm
3
. From (13.10), n
i
= 62.5 particles cm
3
, and from (13.13), a
i
=
4.9 10
7
cm
2
cm
3
.
13.4 CONTINUOUS SIZE DISTRIBUTIONS
Once a discrete model size distribution has been laid out, the initial particle number,
volume, and mass concentrations must be distributed among model size bins. This
can be accomplished by tting measurements to a continuous size distribution,
then discretizing the continuous distribution over the model bins. Three continuous
distributions available for this procedure are the lognormal, MarshallPalmer, and
modied gamma distributions.
13.4.1 The lognormal distribution
A unimodal lognormal distribution is a bell-curve distribution on a log scale, as
shown in Fig. 13.2(a). Figure 13.2(b) shows the same curve on a linear scale. When
two, three, or four lognormal modes are superimposed on each other, the resulting
lognormal distribution is bi-, tri-, or quadrimodal, respectively.
A lognormal distribution has a characteristic geometric-mean diameter
(geometric-mean mass, volume, surface area, or number diameter), a geometric
standard deviation, and a total mass, volume, surface area, or number concentra-
tion associated with it. Fifty percent of the area under a lognormal distribution
lies below the geometric-mean diameter. It is analogous to the mean diameter of
a normal distribution. The geometric standard deviation is dened so that about
454
13.4 Continuous size distributions
10
3
10
2
10
1
10
0
10
1
10
2
0.001 0.01 0.1 1
d
v

(
m
3

c
m
3
)

/

d

l
o
g
1
0
D
p
Particle diameter (D
p
, m)
D
2
D
1
(a)
10
3
10
2
10
1
10
0
10
1
10
2
0 0.05 0.1 0.15
d
v

(
m
3

c
m
3
)

/

d

l
o
g
1
0
D
p
p
, m)
(b)
Figure 13.2 (a) A lognormal particle volume distribution. The volume con-
centration (dv) of material between any two diameters (D
1
and D
2
) is
estimated by multiplying the average value from the curve between the two
diameters by log
10
D
2
log
10
D
1
. (b) The lognormal curve shown in (a),
drawn on a linear scale.
68 percent of the area under the lognormal distribution lies within one geometric
standard deviation of the geometric-mean diameter. About 95 percent of the area
falls within two standard deviations of the geometric-mean diameter. The geomet-
ric standard deviation of a lognormal distribution is analogous to the standard
deviation of a normal distribution.
13.4.1.1 Obtaining parameters for a lognormal distribution
Parameters for lognormal curves may be obtained from measurements. A low-
pressure impactor is an instrument that collects particles of different size. An
impactor with seven stages, for example, collects particles in seven size regimes,
such as 0.050.075, 0.0750.12, 0.120.26, 0.260.5, 0.51.0, 12, and 24 m
in diameter (e.g., Hering et al. 1979).
Suppose particles collected on a seven-stage impactor are weighed and their mass
is plotted. If the plot is somewhat lognormal (e.g., if it looks like Fig. 13.2(a)), the
geometric-mean mass diameter of the size distribution can be found from
ln

D
M
=
1
M
L
I
s
j =1
(m
j
lnd
j
) M
L
=
I
s
j =1
m
j
(13.16)
where I
s
is the number of stages (7 for a seven-stage impactor), M
L
is the total mass
concentration of particles (g m
3
), summed over all impactor stages, and m
j
is
the mass concentration of particles with average diameter d
j
in impactor stage j.
The geometric-mean volume diameter of the distribution is then
ln

D
V
=
1
V
L
I
s
j =1
v
j
lnd
j
V
L
=
I
s
j =1
v
j
v
j
=
m
j
c
m
j
(13.17)
455
10
3
10
1
10
1
10
3
10
5
0.001 0. 01 0. 1 1
d
x
/
d

l
o
g
1
0
D
p

(
x

=

n
,

a
,

v
)
p
, m)
D
V
D
N
D
A
n
v
a
Figure 13.3 Number (n, particles cm
3
), area
(a, cm
2
cm
3
), and volume (v, cm
3
cm
3
) con-
centrations and corresponding geometric-mean
diameters for a lognormal distribution.
where V
L
is the total volume concentration (cm
3
cm
3
) of particles, summed over
all stages, v
j
is the volume concentration of particles in impactor stage j, and
j
is the average density of particles in stage j. If the average density of particles is
constant across all stages,

D
V
=

D
M
.
The geometric-mean surface-area diameter (

D
A
) is always smaller than

D
V
, and
the geometric-mean number diameter (

D
N
) is always smaller than

D
A
(Fig. 13.3).
These latter parameters are given by
ln

D
A
=
1
A
L
I
s
j =1
(a
j
lnd
j
) A
L
=
I
s
j =1
a
j
a
j
=
3m
j
c
m
j
r
j
(13.18)
ln

D
N
=
1
N
L
I
s
j =1
(n
j
lnd
j
) N
L
=
I
s
j =1
n
j
n
j
=
m
j
c
m
j
(13.19)
where A
L
is the total area concentration (cm
2
cm
3
) of particles, summed over all
stages, a
j
is the surface area concentration of particles in stage j, N
L
is the total
number concentration (particles cm
3
) of particles, summed over all stages, and n
j
is the number concentration of particles in stage j.
From the geometric-mean diameters, the geometric standard deviation of the
distribution is calculated with
ln
g
=
1
M
L
I
s
j =1
m
j
ln
2
d
j
D
M
1
V
L
I
s
j =1
v
j
ln
2
d
j
D
V
1
A
L
I
s
j =1
a
j
ln
2
d
j
D
A
1
N
L
I
s
j =1
n
j
ln
2
d
j
D
N
(13.20)
Thus,
g
is the same for a mass, volume, surface-area, or number distribution.
456
13.4.1.2 Creating a model lognormal distribution
Once parameters have been found for a lognormal mode, the mode can be dis-
cretized over a model size grid. The mass concentration of particles discretized
from the continuous lognormal distribution to model size bin i is
m
i
=
M
L
Ld
i
d
i
2 ln
g
exp
ln
2
(d
i

D
M
)
2ln
2
(13.21)
where Ld
i
was dened in (13.9) for the volume-ratio size distribution. Once mass
concentration is known, the volume concentration, surface-area concentration, and
number concentration for a size bin can be calculated from (13.14), (13.13), and
(13.10), respectively. Alternatively, if V
L
, A
L
, or N
L
is known for a lognormal mode,
the volume, surface-area, or number concentration in a size bin can be calculated
from
v
i
=
V
L
Ld
i
d
i
2 ln
g
exp
ln
2
(d
i

D
V
)
2ln
2
(13.22)
a
i
=
A
L
Ld
i
d
i
2 ln
g
exp
ln
2
(d
i

D
A
)
2ln
2
(13.23)
n
i
=
N
L
Ld
i
d
i
2 ln
g
exp
ln
2
(d
i

D
N
)
2ln
2
(13.24)
respectively. Figure 13.3 shows a plot of number, surface area, and volume
concentrations for a lognormal distribution calculated from (13.22)(13.24).
The total volume and number concentrations in a lognormal mode are related
by
V
L
=
0
v
d
dd =

6
0
n
d
d
3
dd =

6
D
3
N
exp
9
2
ln
2
N
L
(13.25)
Thus, if N
L
is known, V
L
and v
i
can be found from (13.25) and (13.22), respec-
tively.
13.4.1.3 Multiple lognormal modes and particle components
Particle distributions in the atmosphere are often described by up to four lognor-
mal modes. Such modes may include a nucleation mode, two subaccumulation
modes, and a coarse-particle mode. The nucleation mode (geometric-mean diame-
ter -0.1 m) contains homogeneously nucleated or small emitted particles. Sulfuric
acid and water homogeneously nucleate together under the right conditions. Hot
organic and inorganic vapors from combustion nucleate quickly to form small
457
particles. Automobiles also emit soot particles (black carbon plus organic matter)
in the nucleation mode. Small nucleated or emitted particles increase in size by
coagulation and growth.
Growth and coagulation move particles into the accumulation mode, where
diameters are 0.12 m. Some of these particles are lost by rainout and washout,
but they are too light to sediment out signicantly. Particles in the nucleation and
accumulation modes, together, are ne particles. The accumulation mode often
consists of two sub-modes with geometric-mean mass diameters near 0.2 m and
0.50.7 m, respectively (Hering and Friedlander 1982; John et al. 1989). Hering
and Friedlander observed that the mass-median diameter of particles containing
sulfate in an urban area was about 0.20 m on dry days ( f
r
= 1768 percent) and
0.54 m on moist days ( f
r
= 26100 percent).
The coarse-particle mode consists of particles larger than 2-m diameter. These
particles originate from wind-blown dust, sea spray, volcanos, plants, and other
sources and are generally heavy enough to sediment out (Whitby 1978). The emis-
sion sources and deposition sinks of ne particles differ from those of coarse par-
ticles. Fine particles usually do not grow by condensation much larger than 1 m,
indicating that coarse-mode particles originate primarily from emission.
The mass of individual components within particles can be described with log-
normal modes different from those of the whole particle. Mylonas et al. (1991)
found that the distribution of organonitrates in ambient aerosol particles is typically
bimodal. Venkataraman and Friedlander (1994) found that ambient black carbon
had two distinct modes, one between 0.05 and 0.12 m and the other between 0.5
and 1 m in diameter. Sulfate in coarse-mode particles has been observed in Los
Angeles to have a larger geometric-mean mass diameter than nitrate (Noll et al.
1990). Sodium and chloride appear mostly in large sea-spray particles. Kritz and
Rancher (1980) found the geometric-mean volume diameter particles containing
NaCl over the ocean to be about 6.9 m.
To initialize a model size distribution with several particle components, each
with multiple lognormal modes, the following steps may be taken. First, lognormal
parameters, such as the total mass concentration M
L.q.k
, geometric-mean mass
diameter

D
M.q.k
, and geometric standard deviation
g.q.k
, are obtained for each
mode k and component q with (13.16) and (13.20). Second, the mass concentration
of each component (m
q.i.k
) is distributed into each size bin i from each lognormal
mode with (13.21). When a highly resolved model size distribution is used (e.g., as
V
rat
approaches unity), (13.21) conserves mass. Thus,
N
B
i =1
m
q.i.k
M
L.q.k
(13.26)
Third, the mass concentration of component q in bin i is summed over all lognormal
modes. When four lognormal modes are used, the summation is
m
q.i
= m
q.i.1
+m
q.i.2
+m
q.i.3
+m
q.i.4
(13.27)
458
10
1
10
0
10
1
10
2
10
3
10
4
10
5
10
6
0
50
100
150
200
250
300
0.01 0.1 1 10
p
, m)
dn (No. cm
3
)
/d log
10

D
p
da (m
2
cm
3
)/d log
10
D
p
dv (m
3
cm
3
)
/d log
10
D
p
Figure 13.4 Number-, area-, and volume-
concentration size distribution of particles at
Claremont, California, during the morning of
August 27, 1987. Sixteen model size bins and
four lognormal modes were used to simulate the
distribution (Jacobson 1997a).
Fourth, the volume concentration of component q in size bin i is calculated from
the summed mass concentration as
v
q.i
=
m
q.i
c
m
q
(13.28)
The number concentration in bin i is then determined from
n
i
=
1
i
N
V
i =1
v
q.i
(13.29)
which was obtained by substituting (13.12) into (13.10).
Figure 13.4 shows a quadrimodal lognormal distribution, tted from data
at Claremont, California, for the morning of August 27, 1987. The particles
contained organic carbon, black carbon, ammonium, nitrate, sodium, chloride,
liquid water, and crustal material (e.g., Mg, Al, Si, K, Ca, and Fe). All four modes
are most noticeable in the number concentration distribution. The nucleation mode
is less noticeable in the surface-area concentration distribution and invisible in the
volume concentration distribution.
Table 13.3 shows trimodal lognormal parameters for a size distribution that
describes typical continental particles. The table shows that the nucleation mode
has the highest particle number concentration, the accumulation mode has the
highest surface-area concentration, and the coarse mode has the highest volume
concentration of particles.
13.4.2 MarshallPalmer distribution
A second type of continuous size distribution, used to parameterize raindrops, is
the MarshallPalmer distribution (Marshall and Palmer 1948). In discretized form,
459
Table 13.3 Lognormal parameters for typical continental aerosol particles
Parameter Nucleation mode Accumulation mode Coarse-particle mode
g
1.7 2.03 2.15
N
L
(particles cm
3
) 7.7 10
4
1.3 10
4
4.2
D
N
(m) 0.013 0.069 0.97
A
L
(m
2
cm
3
) 74 535 41
D
A
(m) 0.023 0.19 3.1
V
L
(m
3
cm
3
) 0.33 22 29
D
V
(m) 0.031 0.31 5.7
Source: Whitby (1978).
the distribution gives the drop number concentration (particles cm
3
) between
diameters d
i
(m) and d
i
+Ld
i
as
n
i
= Ld
i
n
0
e
r
d
i
(13.30)
where Ld
i
n
0
is the value of n
i
at d
i
= 0, and
r
is an empirical parameter that
depends on the rainfall rate. Marshall and Palmer found n
0
= 8 10
6
particles
cm
3
m
1
and
r
= 4.1 10
3
R
0.21
m
1
, where R is the rainfall rate in mil-
limeters per hour. Typical rainfall rates are 125 mm h
1
. If (13.30) is applied to
a volume-ratio size distribution, Ld
i
is found from (13.9). The total number con-
centration (particles cm
3
) and liquid water content (g m
3
) in a MarshallPalmer
distribution are n
T
= n
0
r
and w
L
= 10
6
w
n
0
4
r
, respectively, where
w
is the
liquid-water density (g cm
3
) and 10
6
converts units of the variables dened.
(Pruppacher and Klett 1997).
Example 13.6
Estimate the number concentration of raindrops, in the diameter range d
i
=
1.0 mm to d
i
+Ld
i
= 2.0 mm of a MarshallPalmer distribution, when the
rainfall rate is R = 5 mm h
1
. What are the total number concentration and
liquid water content in the distribution?
SOLUTION
From (13.30), n
i
= 4.3 10
4
particles cm
3
in the given size range. The total
number concentration and liquid water content are thus n
T
= 0.0027 particles
cm
3
and w
L
= 0.34 g m
3
, respectively.
460
Table 13.4 Modied gamma parameters, liquid water content, and total drop number
concentration for several cloud types
Liquid-water
r
c.g
content Number conc.
Cloud type A
g

g

g
( m) (g m
3
) (particles cm
3
)
Stratocumulus base 0.2823 5.0 1.19 5.33 0.141 100
Stratocumulus top 0.19779 2.0 2.46 10.19 0.796 100
Stratus base 0.97923 5.0 1.05 4.70 0.114 100
Stratus top 0.38180 3.0 1.3 6.75 0.379 100
Nimbostratus base 0.080606 5.0 1.24 6.41 0.235 100
Nimbostratus top 1.0969 1.0 2.41 9.67 1.034 100
Cumulus congestus 0.5481 4.0 1.0 6.0 0.297 100
Light rain 4.97 10
8
2.0 0.5 70.0 0.117 0.001
When (13.31) is summed over all sizes, the result equals the value in the last column.
Source: Welch et al. (1980).
13.4.3 Modied gamma distribution
A third type of continuous size distribution is the modied gamma distribution
(Deirmendjian 1969), which is used to approximate cloud drop and raindrop num-
ber concentration as a function of size. In discretized form, the distribution gives
the drop number concentration (particles cm
3
) in size bin i as
n
i
= Lr
i
A
g
r
g
i
exp
r
i
r
c.g
(13.31)
where Lr
i
is the radius width of the bin (m), centered at a mean radius of r
i
(m),
r
c.g
is a critical radius (m), around which the entire size distribution is centered,
and A
g
,
g
, and
g
are parameterized coefcients obtained from measurement.
Table 13.4 shows parameters for several cloud types.
Example 13.7
Use the modied gamma size distribution equation to approximate the number
concentration of drops between 14 and 16 m in radius at the base of a stratus
cloud.
SOLUTION
The mean radius of this size bin is r
i
15 m, and the radius width of the bin
is Lr
i
=2 m. FromTable 13.4 and (13.31), the resulting number concentration
of stratus base drops in the size bin is n
i
= 0.1506 particles cm
3
.
461
13.5 EVOLUTION OF SIZE DISTRIBUTIONS OVER TIME
A size structure is dened here as a size distribution that evolves over time. Five
size structures include the full-stationary, full-moving, quasistationary, hybrid, and
moving-center structures. These are discussed below. In all structures, each size bin
contains n
i
particles per cubic centimeter of air, each particle in a size bin has the
same average single-particle volume (
i
) and composition as each other particle in
the bin, and v
i
= n
i
i
.
13.5.1 Full-stationary structure
In the full-stationary size structure (e.g., Turco et al. 1979; Gelbard and Seinfeld
1980), the average volume of particles in a size bin (
i
) stays constant, but n
i
and
v
i
change throughout a model simulation. During growth,
i
does not increase;
instead, n
i
and v
i
decrease in one size bin and increase in a larger bin. Thus,
particles in a bin grow numerically, not by increasing in volume in their own bin,
but by moving across diameter space to a larger size bin.
An advantage of the full-stationary structure is that it covers a wide size dis-
tribution in diameter space with relatively few size bins. During nucleation, new
particles are placed into an existing size bin, usually the smallest. During emission,
new particles are placed in or partitioned between existing size bins. During coag-
ulation, particles from one size bin collide with and stick to particles from another
bin, forming a larger particle, a fraction of which is partitioned by number between
two bins. During transport, particles in one size bin and grid cell advect and replace
particles with the same
i
in adjacent grid cells. The full-stationary structure is con-
venient for nucleation, emission, coagulation, and transport because
i
is always
constant.
During growth with the full-stationary size structure, particles are advected
in diameter space across size-bin boundaries just as they are advected in real space
by the winds (Chapter 6). During growth (advection) in diameter space, some par-
ticles remain in the same size bin and others move to the next bin, where they
are averaged with particles remaining in the new bin (some of the particles in the
new bin also grow to the next larger bin). Thus, when growth occurs, informa-
tion about the original composition of the growing particle is lost due to aver-
aging, producing errors in the size distribution upon evaporation. For example,
when liquid water condenses onto aerosol particles to form a fog, the original
aerosol material from many small size bins is grown into a few large bins, and
the composition of the material in each large bin is now an average of material
from several small bins. Upon evaporation of the fog, aerosol material does not
remember where it originated from, and evaporation results in a distribution of
core aerosol material different fromthat prior to growth. Figure 13.5 illustrates this
problem.
A second disadvantage of the full-stationary structure is that particle growth
leads to numerical diffusion in diameter space just as Eulerian advection leads to
numerical diffusion in real space. When particles in size bin A grow, the number
of particles in a larger bin (e.g., bin B) increases. The new particles in bin B are
462
13.5 Evolution of size distributions over time
Figure 13.5 Demonstration of a problem
with the full-stationary size structure. Sup-
pose two particles exist in one size bin,
one particle exists in a second size bin,
and all three particles are composed of
the same material. Suppose also that liq-
uid water condenses on all particles, caus-
ing them to grow to the same total size. In
a full-stationary size structure, the three
particles are moved to one size bin with
volume
i
. Since the composition of all
particles in a bin must be the same, the core
volume of each particle must be averaged
with those of all other particles in the bin.
If the water now evaporates, the resulting
distribution differs from the original dis-
tribution although it should not.
assumed to have the same average diameter, d
B
, as all other particles in the bin.
In reality, the new particles in bin B may be smaller than d
B
. During the next time
step, new particles at the average diameter in bin B grow to the next larger bin,
C, where the average diameter is d
C
. In reality, they should have grown from a
diameter smaller than d
B
to a diameter smaller than d
C
and that may not even
be in bin C. The articial spreading of the distribution into bin C is numerical
diffusion.
13.5.2 Full-moving structure
The extreme alternative to the full-stationary size structure is a full-moving size
structure (e.g., Mordy 1959; Gelbard 1990). This structure may be initialized with
a volume ratio or another size distribution. During growth, the number concen-
tration (n
i
) of particles in a size bin does not change. Instead, the single-particle
volume (
i
) changes. Once
i
changes, the average volume of particles in one size
bin no longer equals V
rat
multiplied by the average volume of particles in a previous
size bin and particle volumes no longer necessarily increase with increasing size-bin
number. The full-moving structure is analogous to Lagrangian horizontal advec-
tion, just as the full-stationary size structure is analogous to Eulerian horizontal
advection.
463
Figure 13.6 Preservation of an aerosol distribution upon
growth and evaporation in a full-moving structure. When
water condenses onto particles of different size, all par-
ticles may grow to the same total volume, but the core
within each particle is not averaged with those of other
particles of the same size. Thus, upon evaporation, the
original core distribution is preserved.
A
B
Figure 13.7 Reordering of particle size bins for coag-
ulation in a full-moving structure. After growth, par-
ticles in each bin have new volumes, as illustrated by
row A (the inner circle in each particle is core mate-
rial, and the outer shell is volatile material). During
coagulation, particle bins are reordered from small-
est to largest total volume (row B), and coagulation
proceeds in the same way as with a full-stationary
structure.
One advantage of the full-moving size structure is that core particle material
is preserved during growth. The problem shown in Fig. 13.5 for a full-stationary
structure is not a problem for the full-moving structure, as shown in Fig. 13.6.
The second advantage of the full-moving structure is that it eliminates numerical
diffusion during growth. When one particle grows from d
i
= 0.1 to 0.15 m in
diameter in a full-moving structure, the particle physically attains this diameter
and is not partitioned by volume between adjacent size bins. Particles in a full-
moving structure grow to their exact size, eliminating numerical diffusion. When
two particles coagulate, they form a larger particle whose total volume may be
partitioned between the total volume of particles in two adjacent size bins (Jacobson
1994, 1997a). Thus, coagulation may be treated in the same way as in the full-
stationary structure. During coagulation in the full-moving structure, size bins are
reordered from smallest to largest each time step, since growth changes particle
volume. Figure 13.7 illustrates bin reordering during coagulation.
464
13.5 Evolution of size distributions over time
The full-moving structure has disadvantages. Because it contains a nite number
of size bins, each of which can grow to any volume, it has problems during nucle-
ation, coagulation, and transport. Suppose all particles grow to fog-sized drops
(around 10 m diameter). In such a case, a size bin no longer exists for homo-
geneously nucleated particles. Similarly, if all particles grow to the same large
size, no larger size bin may remain for particles to coagulate into. Transport also
causes problems, since if particles in grid cell X and size bin i grow to volume
X.i
,
and particles in cell Y and bin i grow to volume
Y.i
, then transport of particles
from cell X to Y results in new particles in cell Y with volumes averaged between
X.i
and
Y.i
. Averaging diminishes benets of nondiffusive growth. Because of
the problems described, the full-moving structure is not used in three-dimensional
models.
13.5.3 Quasistationary structure
The quasistationary size structure (Jacobson 1997a) is similar to the full-stationary
structure in most respects. With a quasistationary structure, particle volume grows
to its exact size during one time step, but the adjusted volume is tted back onto
a stationary grid at each step. The simplest way to partition an adjusted volume is
between two adjacent stationary size bins in a number- and volume-concentration-
conserving manner. Suppose that particles in bin i of number concentration n
i
and volume
i
grow by condensation to actual volume
i
which lies between the
volumes of two adjacent xed size bins, j and k (i.e.,
j

i
-
k
). The particles
in bin i can be partitioned between those in bins j and k by solving (1) the number-
conservation relationship
n
i
= Ln
j
+Ln
k
(13.32)
where Ln
j
and Ln
k
are the portions of number concentration n
i
of grown par-
ticles added to size bins j and k, respectively, and (2) the volume-concentration-
conservation relationship
n
i
i
= Ln
j
j
+Ln
k
k
(13.33)
The solution to this set of two equations and two unknowns is
Ln
j
= n
i
j
Ln
k
= n
i
i

j
j
(13.34)
Like the full-stationary structure, the quasistationary structure allows practi-
cal treatment of nucleation, emission, and transport in three dimensions. Like
the full-moving structure, it allows particles to grow to their exact sizes during
a time step. However, because it partitions number and volume concentration after
growth, the quasistationary structure is numerically diffusive and should be used
cautiously.
465
13.5.4 Hybrid structure
The hybrid structure (Jacobson 1994; Jacobson and Turco 1995) is a method that
divides particles into a core and a shell and treats the core with a full-stationary
structure and the shell with a full-moving structure. The core is assumed to consist
of relatively involatile material (e.g., black carbon, soil components, sodium) and
the shell, of relatively volatile material (e.g., liquid water). The average core volume
of particles in a size bin is held constant, but total particle (core plus shell) volume
varies.
When two particles coagulate under this structure, they become one particle
whose core volume (not total volume) is partitioned between two adjacent size bins
in a number- and volume-concentration-conserving manner, such as with (13.34).
During condensation of volatile material, particles grow to their exact sizes, elimi-
nating numerical diffusion, but particles do not move out of the size bin that they
start in since the size bin is controlled by the core volume, not the total volume.
Because particles do not move to a different size bin during growth, their core
volumes are preserved upon evaporation of the shell material.
The hybrid structure has a disadvantage with respect to transport over an Eule-
rian grid. Suppose particles in a small core bin in one grid cell growby condensation
to cloud-sized drops while those in the same bin in an adjacent cell do not. Trans-
port fromone cell to the next results in unrealistic averaging of the volatile material
(but a realistic averaging of the core material). Thus, the hybrid structure, which
is useful for box or Lagrangian modeling, is less useful for Eulerian modeling.
13.5.5 Moving-center structure
The moving-center structure (Jacobson 1997a) is a size structure in which size-
bin edges are xed but size-bin centers vary between the low and high edges of
the bin. Thus, for example,
i,hi
,
i,lo
, and L
i
are xed, and
i
varies between
i,hi
and
i,lo
. Because size-bin edges are xed, homogeneous nucleation, emission,
coagulation, and transport are treated in the same way as in the full-stationary
structure. When growth occurs, though, particles in a size bin are allowed to grow
to their exact size. If the average volume of a particle in a bin grows larger than
the high-edge volume of the bin, all particles in the bin are moved to a single size
bin bounding the average volume.
Because all particles are moved to the same size bin and not fractionated among
two or more size bins, numerical diffusion is reduced with the moving-center struc-
ture. Some diffusion can occur, since the volume of particles moved to the larger (or
smaller) bin is averaged with the volume of particles already in the bin. Figure 13.8
shows, though, that numerical diffusion fromthe moving-center structure was very
small relative to that from the full-moving structure during a case of cloud-drop
growth. The gure also shows that the quasistationary structure was very diffusive.
Zhang et al. (1999) similarly found that the simulation of condensational growth
with the moving-center structure resulted in the accurate predictions among several
techniques compared.
466
13.6 Summary
10
1
10
1
10
3
10
5
10
7
0.1 1 10 100
d
v

(
m
3

c
m
3
)

/
d

l
o
g
1
0

D
p
p
, m)
Initial
Full-moving
Quasi-
stationary
Moving-center
Figure 13.8 Comparison of moving-center,
full-moving, and quasistationary size structure
after growth of water onto aerosol particles
to form cloud-sized drops. Growth occurred
when the relative humidity was increased to
100.002 percent and replenished every 60-s
time step for a 10-min period. Results from the
moving-center and full-moving distributions
were nondiffusive and nearly identical. The
quasistationary structure was very diffusive.
When nucleation or emission occurs with the moving-center structure, new par-
ticles enter the size bin with a high- and low-edge volume surrounding that of
the new particles, and the new and existing particles in the bin are averaged by
volume. When transport occurs, particles in a bin move and replace particles in
adjacent grid cells with the same high- and low-edge volumes. When two particles
coagulate, they form one particle that is partitioned by volume between two adja-
cent size bins. A disadvantage of the moving-center structure is that evaporation
following growth results in a distribution of core aerosol material different from
the initial distribution prior to growth. This is the same problem that occurs with
the full-stationary structure.
In sum, the moving-center structure maintains the main advantages of the full-
stationary structure but also nearly eliminates numerical diffusion during growth.
Because the moving-center structure treats nucleation, emission, coagulation, and
transport realistically, it is more useful than the full-moving structure in three
dimensions.
13.6 SUMMARY
In this chapter, particle composition and size distributions were introduced. Par-
ticles can contain hundreds of components, many of which form from gas-to-
particle conversion and chemical reaction. Size distributions are used to describe
number, surface area, volume, or mass concentrations of particles versus diameter.
Size distributions may be continuous or discrete. Some types of continuous size
distributions include the lognormal, MarshallPalmer, and the modied gamma
467
distributions. A model size structure simulates the change of a discrete size distri-
bution over time. Size structures discussed include the full-stationary, full-moving,
quasistationary, hybrid, and moving-center structures. The full-moving structure
eliminates numerical diffusion during growth, but treats nucleation, emission,
and transport unrealistically and is not used for Eulerian simulations. The full-
stationary and quasistationary structures treat all these processes realistically, but
both lead to numerical diffusion during growth. The hybrid structure eliminates
numerical diffusion during growth but treats transport unrealistically. The moving-
center structure treats nucleation, coagulation, and transport like the full-stationary
structure, but minimizes numerical diffusion during growth. All structures have
advantages and drawbacks.
13.7 PROBLEMS
13.1 Using the diameter information from Table 13.1, calculate how many (a) gas
molecules, (b) medium aerosol particles, and (c) fog drops make up the vol-
ume of a single medium raindrop. Use average diameters where applicable.
13.2 Calculate the high-edge diameter of the largest size bin of a volume-ratio size
distribution with 25 size bins and a volume ratio of adjacent size bins equal
to 2 when the average diameter of the smallest bin is 1 nm.
13.3 If a population of 100-nm diameter particles has a number concentration of
10
5
particles cm
3
-air and a mass density of 1.5 g cm
3
-particle, calculate
the (a) area concentration, (b) volume concentration, and (c) mass concen-
tration of particles. If the same mass concentration were distributed over
particles 500-nm in diameter, what would be the number concentration of
the population?
13.4 Calculate the number concentration of raindrops from the MarshallPalmer
distribution in the diameter ranges (a) 200300 m and (b) 800900 m,
when R = 25 mm h
1
. Why are the number concentrations different in the
two size ranges?
13.5 Compare the number concentrations of drops between 18 and 22 m in
diameter at the base and at the top of a nimbostratus cloud using a modied
gamma distribution. Why do you think the concentrations differ in the two
cases?
13.6 Suppose 10
5
particles cm
3
grow by condensation to a diameter of 250 nm.
If this size falls between two xed size bins of respective diameters 100 nm
and 200 nm, calculate the number concentration of the grown particles par-
titioned to each of the two xed bins with the quasistationary size structure
assuming all particles are partitioned and number concentration and volume
concentration are conserved.
13.7 Write a program to nd
i
,
i.lo
,
i.hi
, d
i
, and dd
i
for a volume-ratio size
distribution when d
1
= 0.005 m, d
N
B
= 500 m, and V
rat
= 1.5. Print a table
of results.
468
13.8 (a) Calculate

D
M
and
g
for the size distribution resulting from the data in
the accompanying table. The data give a typical distribution of emitted
soot particles over a freeway.
Stage Stage diameter Mass between d
j
Stage Stage diameter Mass between d
j
j d
j
(m) and d
j+1
(g m
3
) j d
j
(m) and d
j1
(g m
3
)
8 4 4 0.26 0
7 2 0 3 0.12 2.61
6 1 0 2 0.075 47.1
5 0.5 0 1 0.05 10.5
(b) Using

D
M
and
g
from(a), replot the size distribution on a curve of d
i
ver-
sus dm(M
L
dlog
10
d
p
), where dlog
10
d
p
= dlog
10
d
i
=log
10
d
i+1
log
10
d
i
,
M
L
is the total mass concentration (g m
3
), and dmis the mass concen-
tration in each size bin (subscript i refers to the model distribution size
bin and subscript j refers to the data distribution size bin). To obtain the
curve, use a volume-ratio size distribution with V
rat
= 1.5, and assume
the particles are spherical. Choose appropriate values for d
1
and N
B
.
(c) If soot is emitted primarily in small particles, why is it observed in larger
particles away from emission sources?
469
14
Aerosol emission and nucleation
T
he two processes that increase the number concentration of particles in the
atmosphere are emission and homogeneous nucleation. Horizontal advection,
vertical convection/eddy diffusion, and sedimentation also move particles into
or out of a region. Particle emission originates from natural and anthropogenic
sources. Natural sources include wind uplift of sea spray, soil dust, pollen, spores,
and bacteria, volcanic outgassing, natural biomass res, and lightning. Anthro-
pogenic sources include fossil-fuel combustion, biofuel burning, biomass burning,
and wind uplift of soil dust over eroded land. Globally, particle emission rates
from natural sources exceed those from anthropogenic sources. In urban areas, the
reverse is true. Nucleation may occur on the surface of an existing particle (het-
erogeneous nucleation) or in the absence of pre-existing surfaces (homogeneous
nucleation). Nucleation may also involve the agglomeration of one, two, or more
types of molecules. This chapter discusses aerosol emission and nucleation.
14.1 AEROSOL EMISSION
Emission is a source of primary particles in the atmosphere. In the subsections
below, several types of natural and anthropogenic particle emission sources are
discussed.
14.1.1 Sea-spray emission
The most abundant natural aerosol particles in the atmosphere, in terms of mass,
are sea-spray drops. Sea spray forms when winds and waves force air bubbles
to burst at the sea surface (Woodcock 1953). Sea spray is emitted primarily in the
coarse mode of the particle size distribution. Winds also tear off wave crests to form
larger drops, called spume drops, but these drops fall back to the ocean quickly.
Sea spray initially contains all the components of seawater. About 96.8 percent
of sea-spray weight is water and 3.2 percent is sea salt, most of which is sodium
chloride. Table 14.1 shows the relative composition of the major constituents in
seawater. The chlorine-to-sodium mass ratio in seawater, which can be obtained
from the table, is about 1.8 to 1.
When seawater is emitted as sea spray or spume drops, the chlorine-to-sodium
mass ratio, originally 1.8:1, sometimes decreases because the chlorine is removed
by sea-spray acidication (e.g., Eriksson 1960; Duce 1969; Martens et al. 1973;
Hitchcock et al. 1980; Kritz and Rancher 1980). Sea-spray acidication occurs
470
14.1 Aerosol emission
Table 14.1 Concentrations and mass percentages of major constituents in seawater
Concentration Mass Concentration Mass
Constituent (mg L
1
) percentage Constituent (mg L
1
) percentage
Water 1 000 000 96.78 Sulfur 905 0.0876
Sodium 10 800 1.05 Calcium 412 0.0398
Chlorine 19 400 1.88 Potassium 399 0.0386
Magnesium 1290 0.125 Carbon 28 0.0027
Source: Lide (2003). The total concentration of all constituents is 1 033 234 mg L
1
.
when sulfuric or nitric acid enters a sea-spray drop and forces chloride to evaporate
as hydrochloric acid. Some sea-spray drops lose all of their chloride in the presence
of sulfuric or nitric acid.
The size of a sea-spray drop is also affected by dehydration. Dehydration (loss of
water) occurs when water from a drop evaporates due to a decrease in the relative
humidity between the air just above the ocean surface and that a fewmeters higher.
Dehydration increases the concentration of solute in a drop.
Sea-spray and spume drop emission parameterizations have been developed by
Monahan et al. (1986), Andreas (1992), Wu (1993), Smith and Harrison (1998),
de Leeuw et al. (2000), and Martensson et al. (2003), among others, from mea-
surements of drop concentration as a function of drop size and wind speed. For
example, the parameterization of Monahan et al. (1986) gives the ux of sea-spray
drops, LF
spray
i
(particles cm
2
s
1
), per unit radius interval Lr
i
= Ld
i
2 (cm),
where Ld
i
is dened in (13.9), as
LF
spray
i
Lr
i
= 1.373|v
h.10
|
3.41
(1 +0.057r
1.05
i
)10
1.19e
B
2
r
3
i
(14.1)
where r
i
is the mean radius of the particle (m) at the ambient relative humidity in
the radius interval, |v
h.10
| is the mean horizontal wind speed at 10 m (ms
1
), and
B = (0.38 log
10
r
i
)0.65 (14.2)
The parameterization is valid for 0.8 r
i
10 m and |v
h.10
| - 20 ms
1
. The
corresponding parameterization for spume drops is
LF
spume
i
Lr
i
=
0 r
i
- 10m
8.6 10
6
e
2.08|v
h.10
|
r
2
i
10 r
i
- 75m
4.83 10
2
e
2.08|v
h.10
|
r
4
i
75 r
i
- 100m
4.83 10
6
e
2.08|v
h.10
|
r
8
i
100m r
i
(14.3)
Figure 14.1(a) shows (14.1) and (14.3) as a function of wind speed for 10-m-
radius drops. Below wind speeds of about 9 ms
1
, most of the ux of drops
471
0.1
1
10
100
1000
1 10
F
i
/
r
i

(
p
a
r
t
i
c
l
e
s

c
m
2

c
m
1

s
1
)
10-m wind speed (m s
1
)
Spray (M86)
Spume (M86)
0.1
1
10
100
1000
10
4
10
5
10
6
0.1 1 10 100
Spray (M86)
Spume (M86)
SH98
F
i
/
r
i

(
p
a
r
t
i
c
l
e
s

c
m
2

c
m
1

s
1
)
Drop radius (m)
Figure 14.1 Fluxes of sea-spray and
spume drops per radius interval as a
function of (a) wind speed from (14.1)
and (14.3) when the average radius is
10 m and (b) drop radius from (14.1),
(14.3), and (14.4) when the wind speed
is 10 ms
1
. M86 is Monahan et al.
(1986); SH98 is Smith and Harrison
(1998).
is due to sea-spray generation (bursting of whitecap bubbles). Above 9 ms
1
,
most generation is due to spume-drop generation (mechanical disruption of wave
crests).
Guelle et al. (2001) suggested that the sea-spray parameterization of Monahan
et al. (1986) is accurate for predicting observed concentrations of ocean-generated
aerosol particles up to about 4-mradius. They also suggested that, at higher radii,
472
the parameterization of Smith and Harrison (1998), given as
LF
spray
i
Lr
i
= 0.2|v
h.10
|
3.5
e
1.5ln(r
i
3)
2
+0.0068|v
h.10
|
3
e
ln(r
i
30)
2
(14.4)
(where r
i
is in m and |v
h.10
| is in ms
1
) may be more accurate than Monahan
et al.s sea-spray parameterization. The valid ranges of this parameterization are
1150 mradius and 020 ms
1
wind speed. Figure 14.1(b) shows that the param-
eterization of Smith and Harrison produces results close to those of the spume-drop
parameterization of Monahan et al. (1986) for particles larger than 4 m diameter
when the wind speed is 10 ms
1
, suggesting that spume drops may be important
between 4 and 10 mand the use of the Monahan et al. parameterization of spume
drops in this size range may work just as well as the Smith and Harrison param-
eterization. At higher wind speeds (e.g., 20 ms
1
), though, (14.3) predicts much
greater uxes than does (14.4) at all sizes, so the use of (14.4), which was derived in
part from high-wind-speed data, may prevent unrealistic buildup of ocean aerosol
particles in such cases.
From sea-spray and spume-drop ux expressions, the number concentration of
particles (particles cm
3
) emitted to the lowest model layer of thickness Lz (cm)
during a time step h can be calculated with
Ln
i
=
LF
spray
i
Lr
i
+
LF
spume
i
Lr
i
Lr
i
Lz
h (14.5)
The danger of this equation is that particles in a model are generally emitted into
the middle of the lowest model layer, which may be 1050 m above the surface.
Most spume drops are emitted only a few meters before they fall back to the ocean
surface. Thus, the use of (14.3) without considering the height of emission of spume
drops can lead to a signicant overestimate of spume drop concentrations in the
mid boundary layer, which is why the use of (14.4), which predicts lower spume
drop emission at high wind speeds, may be advantageous.
14.1.2 Soil- and fugitive-dust emission
A major emission source of aerosol particles in the atmosphere is the lifting of
soil dust and road dust by the wind. Soil is the natural, unconsolidated mineral
and organic matter lying above bedrock on the surface of the Earth. A mineral is
a natural, homogeneous, inorganic, solid substance with a characteristic chemical
composition, crystalline structure, color, and hardness. Minerals in soil originate
from the breakdown of rocks. Soil organic matter originates from the decay of
dead plants and animals.
Breakdown of rocks to soil material occurs primarily by physical weathering,
which is the disintegration of rocks and minerals by processes that do not involve
chemical reactions. Disintegration may occur when a stress is applied to a rock,
causing it to break into blocks or sheets of different sizes and ultimately into ne
soil minerals. Stresses arise when rocks are subjected to high pressure by soil
or other rocks lying above. Stresses also arise when rocks freeze then thaw or
473
when saline solutions enter cracks and cause rocks to disintegrate or fracture. Salts
have higher thermal expansion coefcients than do rocks, so when temperatures
increase, salts within rock fractures expand, forcing the rock to open and break
apart. One source of salt for rock disintegration is sea spray transported from
the oceans. Another source is desert salt (from deposits) transported by winds.
Physical weathering of rocks on the Earths surface can occur by their constant
exposure to winds or running water. Rocks also disintegrate by chemical weather-
ing, which is the breakdown of rocks by chemical reaction, often in the presence of
water.
Soil-dust particles consist of the minerals and organic material making up soil.
The Earths crust contains several types of minerals. Pure quartz (SiO
2
(s)) is a
clear, colorless mineral that is resistant to chemical weathering. Feldspars, which
make up at least 50 percent of the rocks on the Earths surface, are by far the
most abundant minerals on Earth. Two common types of feldspars are potas-
sium feldspar (KAlSi
3
O
8
(s)) and plagioclase feldspar (NaAlSi
3
O
3
CaAl
2
Si
2
O
8
(s)).
Hematite (Fe
2
O
3
(s)) is an oxide mineral because it includes a metallic element
bonded with oxygen. The iron within it causes it to appear reddish-brown. Calcite
(CaCO
3
(s)) and dolomite (CaMg(CO
3
)
2
(s)) are carbonate minerals. Gypsum
(CaSO
4
2H
2
O(s)) and epsomite (MgSO
4
7H
2
O(s)) are two of only a handful
of sulfate-containing minerals. Clays are odorous minerals resulting from the
weathering of rocks. They are usually soft, compact, and composed of aggre-
gates of small crystals. Several types of clays include kaolinite, illite, smec-
tite, vermiculite, and chlorite. In addition to containing minerals, soil dust con-
tains organic matter, such as plant litter or animal tissue broken down by
bacteria.
Soil dust is lifted into the air by winds. The extent of lifting depends on the wind
speed and particle mass. Most mass of soil dust lifted into the air is in particles larger
than 1 min diameter; thus, soil-dust particles, in terms of mass, are predominantly
coarse mode particles. Those larger than 10 m fall out quite rapidly, but those
between 1 and 10 m can stay in the air for days to weeks or longer, depending on
the height to which they are originally lifted. Particles 1 m in diameter fall 1 km
in about 328 days, whereas those 10 m in diameter fall 1 km in about 3.6 days.
Although most soil-dust particles are not initially lofted more than 1 km in the
air, many are, and these particles can travel long distances before falling to the
ground.
Source regions of soil dust on a global scale include deserts (e.g., the Sahara
in North Africa, Gobi in Mongolia and northern China, Mojave in southeastern
California) and regions where foliage has been cleared by biomass burning and
plowing. Deserts are continuously expanding due to land clearing and erosion on
desert borders. Up to half of all wind-driven soil-dust emission worldwide may be
anthropogenic (Tegen et al. 1996).
Wind-driven soil-dust emission parameterizations have been developed or
described by Bagnold (1941); Gillette (1974); Greeley and Iversen (1985); Marti-
corena and Bergametti (1995); Shao et al. (1996); Alfaro and Gomes (2001); Shao
(2001), among others. Below, the parameterization of Marticorena and Bergametti
474
(1995) is briey discussed. This treatment assumes that soil grains resting on the
surface experience three major forces: the downward force of gravity, interparticle
cohesion forces, and wind-shear stress on their top surface. The effect of the rst
two forces depends on particle size; that of the third depends on the wind speed
and the presence of roughness elements.
Gravity generally prevents particles larger than about 2000 min diameter from
lifting off the ground. These particles, though, roll or creep along the ground in a
strong wind. Particles 602000 m in diameter may be lifted a few centimeters to
a meter off the ground by the wind before falling back down. The wind may pick
these particles up multiple times and transport them a little further each time. The
transport of soil dust by intermittent, short leaps and bounces off the ground is
called saltation, and the 1-m layer above the soil in which saltation occurs is the
saltation layer. Particles smaller than 60 m in diameter can be lifted above the
saltation layer. These are the ones that enter a model atmosphere. Particles smaller
than 10 m in diameter, though, are the ones most likely to travel long distances.
The total horizontal ux, integrated vertically (g cm
1
s
1
), of soil particles of
all sizes in the saltation layer can be calculated as
G= 2.61
a
g
u
3
N
s
i =1
f
sa.i
1 +
u
t
.i
u
u
t
.i
u
Ld
s.i
for u
> u
t
.i
(14.6)
where
a
is air density (g cm
3
), g is gravity (cm s
2
), u
is the friction wind

speed (cm s
1
), 2.61 is a constant of proportionality determined from wind-tunnel
experiments, N
s
is the number of soil grain size bins considered, Ld
s.i
is the diameter
width of a soil grain size bin (cm) (from(14.9) for a volume-ratio size distribution),
f
sa.i
is the fractional cross-sectional area concentration in soil size bin i among all
soil particles, and u
t
.i
is the threshold friction wind speed (cm s
1
).
The size distribution of soil grains may be described by 13 lognormal modes.
Geometric mean diameters and standard deviations for sand in the rst mode range
from 80 to 200 m and 1.2 to 2.7, respectively (e.g., Marticorena et al. 1997).
Ranges in this mode differ slightly for silt, clay, and gravel. When multiple lognor-
mal soil modes are considered, the fractional cross-sectional area concentration of
all soil grains in a size bin is determined by
f
sa.i
=
N
m
k=1
A
L.s.k
A
L.s.T
a
s.i.k
A
L.s.k
(14.7)
where N
m
is the number of lognormal modes, A
L.s.k
is the total surface-area con-
centration (cm
2
cm
3
) of soil particles in mode k, A
L.s.T
is the total surface-area
concentration summed over all modes, A
L.s.k
A
L.s.T
is the fractional area concen-
tration of soil grains in mode k among all modes (determined from soil data), a
s.i.k
is the surface-area concentration of particles in bin i of mode k, and
a
s.i.k
A
L.s.k
=
Ld
s.i
d
s.i
2 ln
g.s.k
exp
ln
2
(d
s.i

D
A.s.k
)
2ln
2
g.s.k
(14.8)
475
is the fractional area concentration of particles in size bin i in mode k, obtained
from (13.23). In this equation, d
s.i
is the diameter (cm) of soil particles in size bin
i,

D
A.s.k
is the geometric mean cross-sectional area diameter in mode k (equal to
the geometric-mean surface area diameter), and
g.s.k
is the geometric standard
deviation of particles in the mode.
The threshold friction wind speed is the friction wind speed below which no
horizontal soil ux occurs. In (14.6), only terms for which u
> u
t
.i
are treated.
The threshold friction wind speed (cm s
1
) can be calculated with
u
t
.i
=
u
ts
.i
f
eff.i
(14.9)
where
u
ts
.i
=
0.129K
i
1.928Re
0.092
s.i
1
0.03 - Re
s.i
10
0.129K
i
[1 0.0858e
0.0617(Re
s.i
10)
] Re
s.i
> 10
(14.10)
is the threshold friction wind speed over a smooth surface (cm s
1
) and
f
eff.i
= 1
ln
z
0
z
0.s.i
ln
0.35
10 cm
z
0.s.i
0.8
(14.11)
is the ratio of the friction wind speed over a smooth surface to that over the actual
surface. In these equations,
Re
s.i
= 1331d
1.56
s.i
+0.38 (14.12)
is the dimensionless friction Reynolds number, which is the ratio of inertial forces to
friction forces in the saltation layer. The diameter in the equation is in centimeters.
In addition,
K
i
=
s
gd
s.i
1 +
0.006
s
gd
2.5
s.i
(14.13)
is a prefactor (cm s
1
), where
s
is the density of soil (g cm
3
). Finally, z
0
is the
surface roughness length for momentum (cm) over the actual terrain, and
z
0.s.i

d
s.i
30
(14.14)
is the surface roughness length over the terrain as if the terrain consisted of a
smooth soil surface (cm) and where d
s.i
is in cm. Once the horizontal dust ux G
is known, the vertical soil dust ux (g cm
2
s
1
) that escapes the saltation layer is
476
estimated as
LF
soil
vert
= G
s
(14.15)
where
s
10
13.4 f
clay
6
(14.16)
(cm
1
) is the ratio of the vertical to horizontal dust ux. This expression assumes
that the ratio is a function of f
clay
, the mass fraction of soil containing clay, which
has a maximum range of 00.2 for use in this equation.
Finally, the change in the number concentration of soil dust particles (particles
cm
3
) in aerosol size bin (not soil size bin) i in the bottom model layer of thickness
Lz (m) over a time step h (s) due to emission is
Ln
i
= f
em.i
LF
soil
vert
i
h
Lz
(14.17)
where
i
is single-particle volume (cm
3
), and f
em.i
is the mass fraction of the total
vertical mass ux of emitted soil particles partitioned into aerosol bin i. The expres-
sion for f
em.i
is similar to that for f
sa.i
given in (14.7), except that, for f
em.i
, geo-
metric mean mass diameters are used. The geometric mean mass diameter of the
smallest mode is about 610 m. Since only particles -60 m in diameter escape
the saltation layer, the size distribution of vertically emitted particles is smaller than
that of particles moving horizontally in the saltation layer.
Soil dust also enters the air from fugitive sources. Fugitive emission is the emis-
sion of particles during road paving, building construction and demolition, and
operation of agricultural machines, trucks, and passenger vehicles. For example,
tractors kick up soil dust with their treads. Passenger vehicles create turbulence
that resuspends particles resting on paved and unpaved roads.
14.1.3 Volcanic eruptions
Over 1500 volcanos have been active in the last 10 000 years. Between 1975 and
1985, 158 volcanos erupted, some sporadically and others continuously (e.g.,
McClelland et al. 1989; Andres and Kasgnoc 1998). Volcanic eruptions result
from the sudden release of gas dissolved in magma, which contains 1 to 4 percent
gas by mass. Water vapor is the most abundant gas in magma, making up 50 to
80 percent of gas mass. Other important gases include carbon dioxide, sulfur diox-
ide, hydrogen sulde (H
2
S), carbonyl sulde (OCS), and molecular nitrogen. Minor
gases include carbon monoxide, molecular hydrogen, molecular sulfur, hydrochlo-
ric acid, molecular chlorine, and molecular uorine.
Volcanos also emit particles that contain the elements of the Earths mantle.
The most abundantly emitted particle components are silicate minerals (minerals
containing Si). Other components include aluminum, iron, sulfate, nitrate, chloride,
sodium, calcium, magnesium, potassium, black carbon, and organic matter. The
emission rate of sulfate particles is only about 3 percent that of sulfur dioxide gas
(Andres and Kasgnoc 1998). The sizes of emitted volcanic particles range from
477
-0.1 to >100 m in diameter. Particles 100 m in diameter take about an hour
to fall 1 km. The only volcanic particles that survive more than a few months
before falling to the ground are those smaller than 4 m. Such particles require no
fewer than 23 days to fall 1 km. Volcanic particles of all sizes are also removed
by rain.
Volcanic gases, such as carbonyl sulde and sulfur dioxide are sources of new
particles. When OCS is injected volcanically into the stratosphere, some of it pho-
tolyzes. The products react to form sulfur dioxide gas, which oxidizes to form
sulfuric acid gas, which homogeneously nucleates to form new sulfuric acidwater
aerosol particles. A layer of such particles, called the Junge layer, has formed in
the stratosphere by this mechanism (Junge 1961). The average diameter of these
particles is about 0.14 m. Sulfuric acid, formed substantially in this manner in
the stratosphere, is the dominant particle constituent, aside from liquid water, in
the stratosphere and upper troposphere. More than 97 percent of particles in the
lower stratosphere and 91 to 94 percent of particles in the upper troposphere
contain oxygen and sulfur (Sheridan et al. 1994).
14.1.4 Biomass burning and forest res
Biomass burning is the anthropogenic burning of evergreen forests, deciduous
forests, woodlands, grassland, and agricultural land, either to clear land for other
use, stimulate grass growth, manage forest growth, or satisfy a ritual. Forest res
are predominantly natural res set by lightning but also unintentional or arson
res set by humans. About 90 percent of all res worldwide are biomass burning
res; the rest are forest res.
Biomass burning produces gases, such as carbon dioxide, carbon monoxide,
methane, nitrous oxide, oxides of nitrogen, and reactive organic gases. It also pro-
duces particles, such as ash, plant bers, soil dust, inorganic compounds, organic
matter, and soot (e.g., Cooke and Wilson 1996; Liousse et al. 1996; Ferek et al.
1998; Reid et al. 1998; Andreae and Merlet 2001; Bond et al. 2004). Ash is the
primarily inorganic solid or liquid residue left after biomass burning, but may also
contain organic compounds oxidized to different degrees. Organic matter (OM)
consists of carbon- and hydrogen-based compounds and often contains oxygen
(O), nitrogen (N), etc., as well. It often appears white or sandy although it some-
times appears yellow, brown, or reddish due to absorption of ultraviolet radiation
by some organic compounds.
Soot contains spherules of black carbon (BC) (also called elemental or graphitic
carbon) bonded together in an amorphous shape and coated by organic matter.
Pure black carbon contains only carbon atoms. The OM in soot contains aliphatic
hydrocarbons, polycyclic aromatic hydrocarbons (PAHs), and small amounts of
O and N (Chang et al. 1982; Kittelson 1998; Reid and Hobbs 1998; Fang
et al. 1999). BC strongly absorbs visible light and gives soot its blackish color. The
ratio of BC to OM produced in soot depends on temperature. High-temperature
ames produce more BC than OM; low-temperature ames (such as in smoldering
biomass) produce less. Because BC is black, the more BC produced, the blacker the
478
smoke. Biomass burning produces about half of all soot worldwide. Most of the
rest is emitted during fossil-fuel combustion.
Vegetation contains low concentrations of metals, including titanium (Ti), man-
ganese (Mn), zinc (Zn), lead (Pb), cadmium (Cd), copper (Cu), cobalt (Co), anti-
mony (Sb), arsenic (As), nickel (Ni), and chromium(Cr). These substances vaporize
during burning then quickly recondense onto soot or ash particles. Young smoke
also contains K
+
, Ca
2+
, Mg
2+
, Na
+
, NH
+
4
, Cl
, NO

3
, and SO
2
4
. Young smoke
particles are found in the upper-nucleation mode and lower-accumulation mode.
14.1.5 Biofuel burning
Biofuel burning is the anthropogenic burning of wood, grass, dung, and other
biomass for electric power, home heating, and home cooking (e.g., Streets and
Waldhoff 1998, 1999; Andreae and Merlet 2001; Bond et al. 2004). Biofuel burn-
ing releases most of the same gas and aerosol constituents as does biomass burning
but often in different percentages because of different composition of fuel and tem-
perature of burning. Home biofuel burning occurs primarily in developing countries
where electric stoves and gas/electric heaters are uncommon. In developed coun-
tries, most biofuel burning is for power generation at centralized power plants.
14.1.6 Fossil-fuel combustion
Fossil-fuel combustion is another anthropogenic emission source. Major fossil fuels
that produce aerosol particles include coal, oil, natural gas, gasoline, kerosene,
diesel, and jet fuel.
Coal is a combustible brown-to-black carbonaceous sedimentary rock formed
by compaction of partially decomposed plant material. The conversion of plant
material to black coal goes through several stages: unconsolidated brown-black
peat evolves to consolidated, brown-black peat coal, to brown-black lignite coal,
to dark-brown-to-black bituminous (soft) coal, then to black anthracite (hard)
coal.
Oil (or petroleum) is a natural greasy, viscous, combustible liquid (at room
temperature) that is insoluble in water. It forms from the geological-scale decom-
position of plants and animals and is made of hydrocarbon complexes. Oil is found
in the Earths continental and ocean crusts.
Natural gas is a colorless, ammable gas, made primarily of methane but also
of other hydrocarbon gases, that is often found near petroleum deposits. As such,
natural gas is often mined along with petroleum. Gasoline is a volatile mixture of
liquid hydrocarbons derived by rening petroleum. Kerosene is a combustible, oily,
water-white liquid with a strong odor that is distilled from petroleum and used as
a fuel and solvent. Diesel fuel is a combustible liquid distilled from petroleum after
kerosene. Jet fuel is a clear, colorless or yellowish combustible liquid distilled from
kerosene and enhanced with additives.
Particle components emitted during the combustion of fossil fuels include soot
(black carbon and organic matter), organic matter alone, sulfate, metals, and y
479
Table 14.2 Metals present in y ash of different industrial origin
Source Metals present in y ash
Smelters Fe, Cd, Zn
Oil-red power plants V, Ni, Fe
Coal-red power plants Fe, Zn, Pb, V, Mn, Cr, Cu, Ni, As, Co, Cd, Sb, Hg
Municipal waste incineration Zn, Fe, Hg, Pb, Sn, As, Cd, Co, Cu, Mn, Ni, Sb
Open-hearth furnaces at steel mills Fe, Zn, Cr, Cu, Mn, Ni, Pb
Sources: Henry and Knapp 1980; Schroeder et al. 1987; Ghio and Samet 1999; Pooley and
Mille 1999.
ash. Coal combustion results in the emission of submicrometer black carbon,
organic matter, and sulfate and sub/supermicrometer y ash and organic mat-
ter. The y ash consists of oxygen, silicon, aluminum, iron, calcium, and mag-
nesium in the form of quartz, hematite, gypsum, and clays. Combustion in gaso-
line engines usually results in the emission of submicrometer organic matter, some
black carbon, sulfate, and elemental silicon, iron, zinc, and sulfur. Combustion of
diesel fuel results in the emission of these components plus ammonium and greater
amounts of black carbon and organic matter. Diesel-powered vehicles without par-
ticle emission-control devices emit 10 to 100 times more particulate mass than do
gasoline-powered vehicles. Jet fuel combustion emits signicant quantities of black
carbon and sulfate.
Most soot from fossil-fuels originates from diesel-fuel, coal, and jet-fuel com-
bustion (e.g., Cooke and Wilson 1996; Liousse et al. 1996; Cooke et al. 1999; Bond
et al. 2004). Soot emitted during fossil-fuel combustion contains BC covered with
a layer of polycyclic aromatic hydrocarbons and coated by a shell of organic and
inorganic compounds (Steiner et al. 1992). In the case of diesel exhaust, most of the
organic coating is lubricating oil (Kittelson 1998). Most fossil-fuel soot from vehi-
cles is emitted in particles less than 0.2 m in diameter (Venkataraman et al. 1994;
ACEA 1999; Maricq et al. 1999). Thus, fossil-fuel combustion usually results in
the emission of upper nucleation mode and lower accumulation mode particles.
14.1.7 Industrial sources
Many industrial processes involve burning of fossil fuels with metals. As such,
industrial processes emit soot, sulfate, y ash, and metals. Fly ash from industrial
processes usually contains Fe
2
O
3
(s), Fe
3
O
4
(s), Al
2
O
3
(s), SiO
2
(s), and various car-
bonaceous compounds that have been oxidized to different degrees (Greenberg
et al. 1978). Fly ash is emitted primarily in the coarse mode (>2 m in diameter).
Metals are emitted during high-temperature industrial processes, such as waste
incineration, smelting, cement kilning, and power-plant combustion. In such cases,
heavy metals vaporize at high temperatures, then recondense onto soot and y-ash
particles that are emitted simultaneously. Table 14.2 lists some metals present in y
ash of different industrial origin. Of all the metals emitted industrially, iron is by
far the most abundant. Lead is no longer used as an additive in gasoline in nearly
480
every country, but it is still emitted industrially during lead-ore smelting, lead-acid
battery manufacturing, lead-ore crushing, and solid waste disposal.
14.1.8 Miscellaneous sources
Several other particle types are emitted anthropogenically and naturally. Tire-
rubber particles are emitted during the constant erosion of a tire at the tire-road
interface. Such particles are mostly >2 m in diameter, although some are smaller.
Pollen grains are moderately large (10120 min diameter) granules containing
male genetic material released fromowers and carried by the wind to other owers
for fertilization. They originate from trees, plants, grasses, and weeds. Pollen-grain
mass density is about 0.91.1 g cm
3
. Pollen grains are generally released in the
morning during a 214 day period in the spring. A single corn plant can emit
1050 million pollen grains during the period. Of 25 million grains released by
a plant and carried by the wind, about 0.51 percent may stay in the air 500 m
downwind.
Spores are small (26 min diameter) one-celled or multicellular, asexual, repro-
ductive or resting organisms that can develop into an adult on their own when the
environment is favorable. They are usually released by fungi and algae and may
be carried by the wind. Spore mass density is 0.51.5 g cm
3
. Spores are generally
released in the afternoon, when the relative humidity is low. At relative humidities
above 95 percent, spores are generally not released. Because they are smaller than
are pollen grains, spores survive longer and have higher concentrations in the air
than do pollen grains.
Bacteria are unicellular microorganisms found in nearly all environments. Their
largest dimension is an average of 1 m, with a range of 0.1600 m. Viruses are
noncellular organisms that consist of a core of RNA or DNA enclosed by a coat of
protein and/or an outer protective membrane. Their largest dimension ranges from
0.02 to 1 m. Plant debris forms from the bacterial decay of plant matter. Bacteria,
viruses, and plant debris can be lifted by the wind and carried long distances.
A stratosphere source of new particles is meteoric debris. Most meteorites disin-
tegrate before they drop to an altitude of 80 km. Those that reach the stratosphere
contain iron, titanium, and aluminum, among other elements. The net contribu-
tion of meteorites to particles in the stratosphere is small (Sheridan et al. 1994;
Pruppacher and Klett 1997).
14.1.9 Summary of the emission sources of aerosol particles
Table 14.3 summarizes many of the components present in each of the nucleation,
accumulation, and coarse particle modes and their major sources. Processes aside
from emission will be discussed in subsequent sections and chapters.
Table 14.4 shows an estimate of particle component mass emission rates over a
400 150-km region of the Los Angeles basin for a day in August 1987. Emission
rates are given for four size regimes: -1, 12.5, 2.510, and >10 m in diameter.
Of the emission in the table, about 90 percent originated from area sources,
7 percent originated from mobile sources, and 3 percent originated from point
481
Table 14.3 Dominant sources of chemicals in nucleation, accumulation, and coarse
mode particles
Nucleation mode Accumulation mode Coarse mode
Homogeneous nucleation
H
2
O(aq), SO
4
2
, NH
4
+
Fossil-fuel emission
BC, OM, SO
4
2
, Fe, Zn
Sea-spray emission
H
2
O, Na
+
, Ca
2+
, Mg
2+
, K
+
,
Cl
, SO
4
2
, Br
, OM
Fossil-fuel emission
BC, OM, SO
4
2
, Fe, Zn
Biomass-burning emission
BC, OM, K
+
, Na
+
, Ca
2+
,
Mg
2+
, SO
4
2
, NO
3
, Cl
, Fe,
Mn, Zn, Pb, V, Cd, Cu, Co,
Sb, As, Ni, Cr
Soil-dust emission
Si, Al, Fe, Ti, P, Mn, Co, Ni,
Cr, Na
+
, Ca
2+
, Mg
2+
, K
+
,
SO
4
2
, Cl
, CO
3
2
, OM
BC, OM, K
+
, Na
+
, Ca
2+
,
Mg
2+
, SO
4
2
, NO
3
, Cl
, Fe,
Mn, Zn, Pb, V, Cd, Cu, Co,
Sb, As, Ni, Cr
Industrial emission
BC, OM, Fe, Al, S, P, Mn, Zn,
Pb, Ba, Sr, V, Cd, Cu, Co, Hg,
Sb, As, Sn, Ni, Cr, H
2
O,
NH
4
+
, Na
+
, Ca
2+
, K
+
, SO
4
2
,
NO
3
, Cl
, CO
3
2
Ash, OM
Industrial emission
Fly-ash
Tire-particle, pollen, spore,
bacteria, plant debris
emission
Condensation/dissolution
H
2
O(aq), SO
4
2
, NH
4
+
,
OM
H
2
O(aq), SO
4
2
, NH
4
+
,
OM
H
2
O(aq), NO
3
Coagulation from
nucleation mode
Coagulation from smaller
modes
sources. Some of the largest sources of PM
10
(particles smaller than 10 m in
diameter) were fugitive dust (containing silicon, aluminum, and iron) from paved
(32 percent) and unpaved (29.7 percent) roads, and construction/demolition emis-
sion (23.5 percent).
Table 14.4 shows that about 21 tons per day of particulate chloride were emit-
ted in Los Angeles in August 1987. The important sources of land-based chloride
are forest burning, gasoline combustion, agricultural burning, replace burning,
chemical manufacturing, and soil dust (CARB 1988; Saxena et al. 1993). Almost all
(98 percent) of anthropogenic hydrochloric acid gas emission is from coal
combustion; much of the rest is from waste incineration (Saxena et al. 1993).
Table 14.4 shows that only 3 tons of particulate nitrate were emitted per day
in 1987 in Los Angeles. The major atmospheric source of particulate nitrate is not
emission but gas-to-particle conversion of nitric acid gas. The table also shows that
about 28 tons per day of black carbon -1 m in diameter were emitted. Much of
this originated from automobile exhaust.
Globally, natural sources contribute more particle mass to the air than do anthro-
pogenic sources. Natural sources emit 2501610 Tg/yr of primary particles and
enough gases to produce 3452080 Tg/yr of secondary-particle material. Anthro-
pogenic sources emit 6224 Tg/yr of primary particles and enough gases to pro-
duce 140396 Tg/yr of secondary-particle material (Jaenicke 1988; Pruppacher
and Klett 1997; Wolf and Hidy 1997). The relative contribution of anthropogenic
sources to global particle loading is increasing in many countries. In urban regions,
anthropogenic particle emission exceeds natural emission.
482
Table 14.4 Emission rates of aerosol constituents in the Los Angeles basin as a function of
particle diameter
Emission rate (kg day
1
)
Substance -1 m 12.5 m 2.510 m >10 m All sizes Total %
Other (O, H, etc.) 147 884 86 431 380 221 712 049 1 326 585 53.167
Silicon 37 312 37 086 183 641 166 527 424 566 17.015
Organic carbon 28 462 9 363 69 967 58 111 165 903 6.649
Aluminum 14 550 14 644 67 991 58 216 155 401 6.228
Iron 7 090 7 189 37 210 38 947 90 436 3.624
Calcium 5 587 5 511 32 619 34 028 77 745 3.116
Sulfates 45 922 894 3 998 3 122 53 936 2.162
Potassium 7 364 3 586 16 266 18 989 46 205 1.852
Black carbon 28 467 1 095 7 247 7 429 44 238 1.773
Unknown 9 919 6 745 11 110 11 903 39 677 1.590
Chloride 11 318 814 4 535 4 796 21 463 0.860
Titanium 1 048 877 4 241 4 716 10 882 0.436
Sulfur 618 573 3 216 2 129 6 536 0.262
Carbonate ion 306 162 2 514 1 879 4 861 0.195
Sodium 569 233 2 080 1 916 4 798 0.192
Manganese 899 521 1 511 1 824 4 755 0.191
Phosphorus 130 286 1 660 1 148 3 224 0.129
Nitrates 1 237 147 935 782 3 101 0.124
Zinc 226 154 729 674 1 783 0.071
Lead 173 156 758 653 1 740 0.070
Barium 79 88 544 856 1 567 0.063
Ammonium 841 51 120 136 1 148 0.046
Strontium 25 42 308 364 739 0.030
Vanadium 94 66 274 280 714 0.029
Copper 132 60 203 208 603 0.024
Cobalt 127 52 158 212 549 0.022
Nickel 130 48 132 158 468 0.019
Chromium 87 26 158 176 447 0.018
Rubidium 11 12 91 100 214 0.009
Zirconium 6 10 80 110 206 0.008
Lanthanum 26 7 52 65 150 0.006
Bromine 68 5 30 24 127 0.005
Arsenic 26 3 10 19 58 0.002
Cadmium 9 2 23 16 50 0.002
Antimony 5 3 15 23 46 0.002
Yttrium 2 3 17 22 44 0.002
Tin 6 7 15 14 42 0.002
Indium 5 1 11 12 29 0.001
Mercury 2 1 11 13 27 0.001
Molybdenum 3 1 7 9 20 -0.001
Silver 5 2 6 6 19 -0.001
Palladium 3 1 5 10 19 -0.001
Selenium 3 0 1 2 6 -0.001
Gallium 0 0 5 0 5 -0.001
Totals 350 776 176 958 834 725 1 132 673 2 495 132 100.00
% of total 14.06 7.09 33.45 45.40 100.00
Emission rates of several species may be signicantly overestimated. The component other com-
prises oxygen, hydrogen, and other components not included in the aerosol analysis (e.g., the oxygen
atoms in oxides of silicon, aluminum, and iron). Data were provided by Allen and Wagner (1992).
483
14.2 NUCLEATION
Nucleation is a process by which gas molecules aggregate to form a cluster that
condenses to form a small liquid drop. If the radius of a cluster reaches a critical
size, the cluster becomes stable and can grow further by condensation. Nucle-
ation is either homogeneous or heterogeneous and either homomolecular, binary,
or ternary. Homogeneous nucleation occurs when one or more gases nucleates
without the aid of a pre-existing surface. Heterogeneous nucleation occurs when
one or more gases nucleates on a pre-existing surface. Homomolecular nucleation
occurs when a single gas, such as a high-molecular-weight condensable organic
gas, nucleates. Binary nucleation occurs when two gases, such as sulfuric acid and
water, nucleate, and ternary nucleation occurs when three gases, such as sulfuric
acid, water, and ammonia, nucleate. In the sections below, nucleation parameteri-
zations are discussed.
14.2.1 Homogeneous nucleation rates from classical theory
Equations based on classical nucleation theory are used to estimate homogeneous
and heterogeneous nucleation rates. Although other theories exist and classical
theory has been criticized for inaccurately simulating nucleation in the atmosphere,
other theories are often too computationally intensive to be used in a large model.
With respect to homogeneous nucleation, classical theory predicts a production
rate of new particles. The rate is derived from
LG = 4r
2
p
4
3
r
3
p
p
R
T
m
q
ln S
q
(14.18)
which states that the change in Gibbs free energy (LG) (J) during nucleation of a
cluster of molecules is affected primarily by surface tension and the saturation ratio
of the gas over the cluster surface. In the equation, r
p
is the radius of the nucleating
cluster (cm),
p
is the mass density of the condensed cluster (g cm
3
), R
*
is the gas
constant (8.31451 J mol
1
K
1
), m
q
is the molecular weight of the condensing gas
(g mol
1
),
p
is the surface tension of the condensed cluster (J cm
2
), and S
q
is
the saturation ratio of the gas. The saturation ratio of a gas is its partial pressure
divided by its saturation vapor pressure over the surface (S
q
= p
q
p
q.s
).
Surface tension is a force per unit distance that tends to minimize the surface
area of a body. The force arises because molecules in a three-dimensional body are
equally attracted in all directions by other molecules, except at the surface, where
a net inward attraction exists. Work needs to be done to increase the surface area
against the surface-tension force arising from the inward attraction. The greater
the tension, the greater the work, or surface energy, required to bring molecules to
the surface from within the body.
Surface tension acts parallel to the boundary surface and is expressed in units
of energy per unit area, which simplies to force per unit distance. An expression
484
14.2 Nucleation
70
75
80
85
90
40 30 20 10 0 10 20 30 40
S
u
r
f
a
c
e

t
e
n
s
i
o
n

(
d
y
n

c
m
1
)
Temperature (C)
Figure 14.2 Surface tension of liq-
uid water against air, obtained from
(14.19).
for the surface tension (dyn cm
1
) of liquid water against air for the temperature
range 40 to 40

C is
wa
=
p
=
n=0
a
n
T
n
c
40 T
c
- 0
76.1 0.155T
c
0 T
c
- 40
(14.19)
(Dorsch and Hacker 1951; Gittens 1969; Pruppacher and Klett 1997), where T
c
is
the temperature in degrees Celsius, a
0
= 75.93, a
1
= 0.115, a
2
= 0.068 18, a
3
=
6.511 10
3
, a
4
= 2.933 10
4
, a
5
= 6.283 10
6
, and a
6
= 5.285 10
8
.
Figure 14.2 shows a plot of (14.19). The gure shows that surface tension decreases
with increasing temperature. It should also theoretically decrease with decreasing
radius of curvature, but researchers have found that the surface tension of a liquid
water cluster with 13 molecules is anywhere from 0 to 40 percent less than that of
a at liquid water surface (Pruppacher and Klett 1997).
Figure 14.3 illustrates the saturation-ratio and surface-tension terms on the right
side of (14.18). When molecules are added to a cluster, they condense, releasing
latent heat. The decrease in the change in energy with increasing particle radius
depicted by the saturation-ratio term shown in the gure represents the energy
available from latent-heat release.
The addition of molecules to the cluster increases the energy required to expand
the surface area of the cluster against the force of surface tension. The surface-
tension term in Fig. 14.3 represents the energy required to expand the surface.
When the sum of the saturation-ratio and surface-tension terms is increasing,
the addition of a molecule to the cluster consumes more energy than it adds, and
the cluster becomes more unstable. When the sum is decreasing, the addition of a
molecule adds more energy than it consumes, and the cluster becomes more stable.
When the sumreaches a maximum, the addition of a molecule causes no net energy
change. The radius at which the maximum occurs is the critical radius. Nucleation
occurs if a cluster contains enough molecules for the cluster radius to exceed the
critical radius. At larger radii, condensation occurs readily on the nucleated cluster.
At smaller radii, the cluster is unstable and may evaporate.
485
410
14
210
14
0
210
14
410
14
410
6
810
6
1.210
5
1.610
5
210
5
G

(
J
)
Particle radius (cm)
Surface-tension
term
Saturation-ratio
term
Sum
r
c
Figure 14.3 The net change in free energy
as a function of radius is the sum of the
surface-tension and the supersaturation terms, as
described in the text. The critical radius is the
radius at which the net change in free energy with
radius is maximized. Conditions for this curve are
T = 288 K,
p
= 7.5 10
6
J cm
2
, S = 1.01,
p
= 1.0 g cm
3
, m
q
= 18 g mol
1
(water).
The critical radius can be found by setting dLGdr
p
=0 in (14.18). The result is
dLG
dr
p
= 8r
p
p
4r
2
p
p
R
T
m
q
ln S
q
= 0 (14.20)
When dLGdr
p
is positive, the cluster is unstable; when it is zero or negative, the
cluster is stable. Solving (14.20) for the radius gives the critical radius as
r
c
=
2
p
m
q
p
R
T ln S
q
(14.21)
Dividing the volume of a critical cluster (4r
3
c
3) by the volume of one gas molecule
[m
q
(
p
A)], where A is Avogadros number (molec. mol
1
), gives the number of
molecules in a nucleated cluster as
n
c
=
32
3
p
m
2
q
A
3
2
p
(R
T ln S
q
)
3
(14.22)
Table 14.5 shows critical radii and number of water molecules in a nucleated cluster
for several saturation ratios.
An expression for the homogeneous homomolecular nucleation rate of particles
(particles cm
3
s
1
) from classical theory is
J
hom
= 4r
2
c
x
Z
n
N
x
exp
LG
hom
k
B
T
(14.23)
where
LG
hom
= 4r
2
c
p
3 (14.24)
486
14.2 Nucleation
Table 14.5 Critical radii and number of water molecules in a
critical cluster
Saturation ratio S Critical radius (m) Number of molecules
1
1.01 0.11 2.03 10
8
1.02 0.055 2.58 10
7
1.10 0.011 2.32 10
5
1.5 0.0028 3010
2 0.0016 603
5 0.0007 48
10 0.00048 16
T = 288 K,
p
= 7.5 10
6
J cm
2
,
p
= 1.0 g cm
3
, and m
q
=
18 g mol
1
.
(J) is the critical change in free energy, which is found by combining (14.18) with
(14.20), k
B
23
J K
1
), N
x
is the number
concentration (molec. cm
3
) of nucleating gas x,
x
= N
x
k
B
T
2

M
x
(14.25)
is the number of gas molecules striking a unit surface area per second (molec.
cm
2
s
1
) (where

M
x
= m
x
/A is the mass, in grams, of one molecule of x
and k
B
=1.380 658 10
16
g cm
2
s
2
K
1
molec.
1
in this expression), and
N
x
exp(LG
hom
k
B
T) is an equilibrium number concentration (particles cm
3
) of
nucleated clusters of critical radius r
c
.
The equilibriumcluster concentration is an ideal cluster number per unit volume
of air that exists when the number of molecules added to a cluster equals the number
lost from the cluster. At theoretical equilibrium, the number of nucleated clusters
of radius r
c
does not change. In homogeneous nucleation theory, the system is
not always in equilibrium, so some clusters form and others evaporate over time,
changing the number of nucleated clusters with time. Equation (14.23) accounts
for the overall rate of change of cluster concentration through the term
x
and
through the Zeldovich nonequilibrium factor (Zeldovich 1942),
Z
n
=
M
x
2r
2
c

p
k
B
T
(14.26)
which takes account of the remaining differences between equilibriumand nonequi-
librium cluster concentration. Its value is less than unity; thus, time-dependent
nucleation results in a cluster concentration below the equilibrium level (Prup-
pacher and Klett 1997).
As illustrated in Example 14.1, water vapor, alone, does not homogeneously
nucleate in the atmosphere, even when its saturation ratio is near its highest
observed level, 1.02. Under such conditions, the critical radius, and thus the
487
critical change in free energy, is too large, causing the exponent in (14.23) to become
zero. Water can only homogeneously nucleate if the critical radius decreases to
-1 nm, which occurs when the saturation ratio is 5 (the relative humidity is 400 per-
cent). Water vapor readily nucleates heterogeneously, since the free energy change
required to nucleate heterogeneously is much less than that required to nucleate
homogeneously. The free energy change required for heterogeneous nucleation is
discussed shortly.
Example 14.1
Calculate the homogeneous nucleation rate of water vapor when its saturation
ratio is 1.02 and T = 288 K.
SOLUTION
From Table 14.5, the critical radius for new-particle formation with this sat-
uration ratio is 0.055 m. Substituting the critical radius into (14.24) with
a surface tension of 73.8 dyn cm
1
(7.38 10
6
J cm
2
) from (14.19) gives
the critical change in free energy as 9.35 10
16
J. Substituting this into the
exponential term of (14.23) with T = 288 K gives the argument of the expo-
nential term as 2.35 10
5
. The exponent of this number gives zero. Thus,
water vapor cannot homogeneously nucleate in the atmosphere under even its
highest observed saturation ratio.
In (14.23), small changes in free energy affect the nucleation rate exponentially.
Changes in the prefactor affect the rate linearly. As such, changes in the free energy
have a much greater effect on the nucleation rate than do changes in the prefactor
term. Because the saturation ratio, which affects the free energy change, is dif-
cult to calculate precisely in the atmosphere, model-predicted free energy changes
are often slightly off. Since small errors in the free energy are amplied in the
exponential term of (14.23), they result in nucleation rate predictions from clas-
sical theory that are often several orders of magnitude off when applied to the
atmosphere.
Classical homogeneous nucleation theory can be extended to systems in which
two gases nucleate. The resulting homogeneous binary nucleation rate is
J
hom
= 4r
2
c
y
N
x
exp
LG
hom
k
B
T
(14.27)
(particles cm
3
s
1
) where the two gases, x and y, are assumed to nucleate together.
Gas x is assumed to have a much larger concentration than gas y (N
x
N
y
). In the
atmosphere, sulfuric acid commonly nucleates homogeneously with water vapor,
which has a much higher number concentration than does sulfuric acid. For sulfuric
acidwater binary nucleation, gas x is water vapor and gas y is sulfuric acid.
488
14.2 Nucleation
Embryo
q
c
Substrate
r
c
R
h
Figure 14.4 For-
mation of a critical
embryo on the surface
of an existing particle.
c
is the contact angle,
R
h
is the radius of the
particle, and r
c
is the
critical radius.
In equilibrium, the number concentration of nucleated clusters of critical radius,
r
c
, is about (N
x
+ N
y
) exp(LG
hom
k
B
T). Since N
x
N
y
, this expression reduces
to N
x
exp(LG
hom
k
B
T), which is used in (14.27). Because the ux of molecules to
a nucleating cluster is limited by the ux of the gas with the lowest concentration,
is written in terms of gas y in (14.27). The equation for
y
(molec. cm
2
s
1
) is
y
= N
y
k
B
T
2

M
y
(14.28)
where

M
y
is the mass of one molecule of gas y.
For basic homogeneous binary nucleation calculations, (14.27) does not include
a Zeldovich nonequilibrium factor. The saturation ratio, surface tension, and par-
ticle density used to calculate LG
hom
must be estimated as weighted averages
between the two nucleating gases (e.g., Zhao and Turco 1995; Kulmala et al. 1998;
Vehkamaki et al. 2002).
For homogeneous ternary nucleation among gases x, y, and z, (e.g., water, sulfu-
ric acid, and ammonia, respectively) where N
x
N
y
N
z
, Equation (14.27) can
be modied by replacing
y
with
y
+
z
, since the ux of the lower-concentrated
gases limits the rate of nucleation. In addition, the surface tension, cluster density,
and saturation ratio must be weighted by the relative composition of the cluster.
Ternary-nucleation parameterizations have been developed for the sulfuric acid
waterammonia system (e.g., Coffman and Hegg 1995; Napari et al. 2002), the
hydrochloric acidwaterammonia system (e.g., Arstila et al. 1999), and the sulfu-
ric acidwatermethanesulfonic acid system(e.g., Van Dingenen and Raes 1993).
14.2.2 Heterogeneous nucleation rates from classical theory
During heterogeneous nucleation, a critical embryo forms on the surface of an
existing particle, as shown in Fig. 14.4. The ability of the embryo to survive breakup
489
0
0.5
1
1.5
2
2.5
0.1 1 10 100
f
h
x
h
= R
h
/ r
c
0
30
60
90
120
150
q = 180
c
Figure 14.5 Values of Fletchers correction
factor as a function of x
h
for seven different
contact angles
c
.
on the surface depends on the contact angle
c
, which is the angle at which an
embryo contacts a substrate at its surface. If the contact angle is 0
, the surface is
completely covered, or wetted, with the embryo. If the contact angle is 180
, the
surface is nonwettable, and no embryo forms. If water readily wets a surface, the
surface is hydrophilic. Otherwise, it is hydrophobic. Contact angles of water vapor
vary with surface composition. The contact angle of water over sand is 4352
,
over soil is 65.268.9
, and over silver iodide is 917

The contact angle of water over a substrate is calculated from Youngs relation,
c
= cos
1
S.a
S.w
w.a
(14.29)
where
S.a
,
S.w
, and
w.a
are the surface tensions of the substrate against air, the
substrate against water, and water against air, respectively.
The free energy change required to heterogeneously nucleate an embryo on a
surface is approximated with
LG
het
= LG
hom
f
h
(x
h
. m
h
) (14.30)
where f
h
is a correction factor that depends on x
h
= R
h
r
c
, the ratio of the radius of
the host particle to the critical radius, and on m
h
= cos
c
, the cosine of the contact
angle. Fletcher (1958) parameterized the correction factor as
f
h
(x
h
. m
h
) = 1 +
1 m
h
x
h
g
h
3
+ x
3
h
2 3
x
h
m
h
g
h
x
h
m
h
g
h
+3m
h
x
2
h
x
h
m
h
g
h
1
(14.31)
where g
h
=
1 + x
2
h
2m
h
x
h
. Figure 14.5 shows a plot of f
h
versus x
h
and
c
.
When
c
= 0
, f
h
and LG
het
are minimized. When
c
=180
, f
h
and LG
het
are maximized. The factors f
h
and LG
het
can have small values even when
x
h
- 1, indicating that clusters may form when the host particle is smaller than the
nucleating cluster.
490
14.2 Nucleation
A classical-theory expression for the heterogeneous nucleation rate (number of
embryos cm
2
s
1
) of new particles on a surface is
J
het
= 4r
2
c
x
exp
LG
het
k
B
T
(14.32)
where
=
0
exp
E
R
(14.33)
is the characteristic time (s) that a gas molecule spends on the surface before bounc-
ing off. In this expression,
0
is the characteristic time of adsorption (s), which is the
inverse of the characteristic frequency of vibration of the molecules, and depends on
the nature of the surface and of the condensing gas(es). The term E is the enthalpy
of adsorption (J mol
1
), which depends on the surface type and properties of
the condensed gas. Hamill et al. (1982) calculated
0
2.4 10
16
s and E
45 188 J mol
1
for water nucleating on solid carbon particles. Korhonen et al.
(2003) calculated
0
2.55 10
13
s and E 44 970 J mol
1
for liquid water,
0
8.62 10
13
1
for sulfuric acid, and
0
2.55
10
13
1
for ammonia nucleating on soil particles.
In (14.32),
x
and
y
are the numbers of molecules of gases x and y, respec-
tively, striking the substrate surface per second (molec. cm
2
s
1
) and can be found
from (14.25) and (14.28). If gas x differs from gas y, the nucleation is heteroge-
neous binary. If the two are the same, it is heterogeneous homomolecular. Common
binary-nucleating gases are sulfuric acid and water vapor, since sulfuric acid has
a low saturation vapor pressure, and water readily equilibrates with sulfuric acid
when sulfuric acid condenses.
J
het
gives the rate per unit surface area at which embryos form when molecules
of gas y transfer to a surface at rate
y
and combine with molecules of x adsorbed
to the surface. The product
x
n
m
is the number of molecules of gas x that
adsorb to a unit area of the particle surface (molec. cm
2
), where n
m
10
15
sites
cm
2
is the maximum number of adsorption sites available per unit area of surface
(Section 11.3.6.3). Species x is chosen as the chemical that limits surface coverage
because, in the case of binary nucleation, it is assumed to be the more highly
concentrated species.
14.2.3 Parameterized homogeneous nucleation rate
The most important homogeneous nucleation process in the free atmosphere is
binary nucleation of sulfuric acid and water. Homogeneously nucleated sulfuric
acidwater particles are typically 320 nm in diameter. In the remote atmosphere
(e.g., over the ocean), homogeneous nucleation events can produce >10
4
particles
cm
3
in this size range.
491
Because classical theory is often unreliable for predicting nucleation rates in
the atmosphere, nucleation parameterizations have been developed to replace
classical theory. Zhang et al. (1999) have compared several such parameteri-
zations. One empirical parameterization, valid for a temperature range 230
300 K, a relative humidity range 0.01100 percent, and a sulfuric-acid concen-
tration range of 10
4
10
11
molec. cm
3
, is given in Vehkamaki et al. (2002).
A two-parameter parameterization, based on calculations of binary nucleation
of sulfuric acid and water for the remote boundary layer by Jaecker-Voirol and
Mirabel (1989), is
J
hom
= 10
7.0(64.24+4.7 f
r
)+(6.13+1.95 f
r
) log
10
N
H
2
SO
4
(14.34)
(Pandis et al. 1994; Russell et al. 1994), where J
hom
is the rate of production of
new sulfuric acidwater particles (particles cm
3
s
1
), f
r
is the relative humidity
(expressed as a fraction), and N
H
2
SO
4
is the number concentration of gas-phase
sulfuric acid molecules (molec. cm
3
). This parameterization is independent of
temperature, so it cannot be used for the upper troposphere or stratosphere.
Example 14.2
Estimate the homogeneous nucleation rate of sulfuric acidwater particles
over the remote ocean when f
r
= 0.9 and when H
2
SO
4
(g) concentrations are
(a) 0.005 and (b) 0.05 g m
3
.
SOLUTION
In the rst case, the number concentration of gas-phase sulfuric acid molecules
is
N
H
2
SO
4
=
0.005g
m
3
m
3
10
6
cm
3
mol
98 10
6
g
6.02 10
23
molec.
mol
= 3.1 10
7
molec.
cm
3
Substituting this result into (14.34) gives J
hom
= 0.004 particles cm
3
s
1
. In
the second case, J
hom
= 3.1 10
5
particles cm
3
s
1
. A factor-of-10 increase
in N
H
2
SO
4
between the two cases increased the nucleation rate by eight orders
of magnitude. In reality, H
2
SO
4
(g) condenses on particles as they nucleate,
reducing the amount of gas available for further nucleation.
14.3 SUMMARY
In this chapter, aerosol emission and nucleation were discussed. Emissions origi-
nate from human and natural sources. In urban regions, human emission domi-
nates. Globally, natural sources dominate. Some natural emission sources include
volcanos and sea-spray emission. Homogeneous nucleation is a second means of
newparticle production. Such nucleation can be homomolecular, binary, or ternary.
492
Homogeneous and heterogeneous nucleation rates can be estimated with classical
nucleation theory, but this theory may lead to large errors in nucleation rates when
applied to the atmosphere. Empirical parameterizations of nucleation rates may be
better in some cases. Once nucleated, small particles collide and coalesce during
coagulation, discussed in Chapter 15.
14.4 PROBLEMS
14.1 Calculate the average ux of sea-spray drops (particles cm
2
s
1
) between
0.8 and 0.9 m in radius when the wind speed at 10 m is 10 m s
1
.
14.2 Calculate the threshold friction wind speed for soil particles 30 m in diam-
eter when the average surface roughness over the overall terrain is 0.001 m,
soil density is 2.5 g cm
3
, and air density is 0.001225 g cm
3
.
14.3 Given identical vapor concentrations of water at both locations, would you
expect more particles to nucleate heterogeneously over the ocean or over
land? Why? Once particles have nucleated, would you expect the particles
to grow larger over the ocean or over land? Why?
14.4 If the gas concentrations of sulfuric acid and nitric acid are the same, why
should sulfuric acid and water homogeneously nucleate more readily than
nitric acid and water?
14.5 (a) Calculate the critical radius of a homogeneously nucleating water drop
when the saturation ratio is (i) 4.0 and (ii) 1.1. Assume
p
= 1.0 g cm
3
and T = 298.15 K in both cases.
(b) Using the information from (a), calculate the classical-theory homoge-
neous nucleation rate of liquid water drops in both cases.
(c) If the air temperature in part (a) (i) decreased from 298.15 to 283.15 K,
what would the new nucleation rate be? (Be sure to calculate the change
in saturation ratio.)
14.6 For the two cases in Problem 14.5(a), calculate heterogeneous nucleation
rates when
c
= 90
, R
h
= 0.5 m,
0
= 2.4 10
16
s, E = 45 188 J mol
1
,
and T = 298.15 K.
14.7 Write a computer script to calculate r
c
, J
hom
, and J
het
from classical nucle-
ation theory for water. Perform calculations at T = 298 K for f
r
= 100400
percent and at f
r
= 400 percent for T = 233315 K. Plot the results. Use
(2.62) to calculate p
v.s
. Use information from Problems 14.5 and 14.6 where
needed.
493
15
Coagulation
C
oagulati on occurs when two particles collide and stick together (coalesce),
reducing the number concentration but conserving the volume concentration
of particles in the atmosphere. Coagulation is a mechanism by which externally
mixed particles become internally mixed. More small particles than large particles
are lost to larger sizes by coagulation, because the atmosphere contains many more
small particles than large particles (Table 13.1). Simulating coagulation in a model
is important, since if coagulation is neglected, erroneously large aerosol number
concentrations will be predicted. Even if the total aerosol mass concentration in
a model is correct, the mass concentration will be spread among too many par-
ticles. Coagulation is also important for determining the number concentration
of particles that exceed the critical radius required for cloud-drop activation and
determining the evolution of nucleated and newly emitted particles. In fogs and
many clouds, coagulation (collision plus coalescence) is the main mechanism by
which cloud drops evolve into raindrops. To simulate coagulation in an Eulerian
model, a numerical scheme must be fast and accurate. One scheme is described
in this chapter. A solution for homogeneous particles over a single size distribu-
tion is rst derived. The solution is then expanded to include any number of size
distributions and particles of any composition.
15.1 IMPLICIT COAGULATION
The derivation of numerical methods for solving coagulation begins with the inte-
grodifferential coagulation equation (Muller 1928)
n
t
=
1
2
.
n

n

d n
.
n

d (15.1)
where and are the volumes of two coagulating particles, is the volume
of the new, coagulated particle, n is the time-dependent number concentration
(particles cm
3
) of particles of volume , , or , and is the coagulation kernel
(rate coefcient) of the two colliding particles (cm
3
particle
1
s
1
). Equation (15.1)
states that the change in number concentration of particles of volume equals the
rate at which particles of volume coagulate with particles of volume minus
494
15.1 Implicit coagulation
v
1
2v
1
3v
1
4v
1
k = 1 k = 2 k = 3 k = 4
Figure 15.1 A monomer size distribution.
The volume of particles in any size bin k
equals the volume of particles in the small-
est bin multiplied by k.
the rate at which particles of volume are lost due to coagulation with particles
of all sizes. The rst integral in (15.1) is multiplied by one-half to eliminate double
counting of production terms.
A few analytical solutions to (15.1) exist. Some are described in Section 15.3.
All analytical solutions require a particular form for the initial aerosol size distri-
bution. Other solutions to the equation are numerical. Some numerical solutions
are based on the assumption that the aerosol size distribution is represented by
lognormal modes (e.g., Pratsinis 1988; Binkowski and Shankar 1995; Binkowski
et al. 2003). Others are based on the assumption that the aerosol distribution is
discrete. Solutions to the discrete form of the coagulation equation are discussed
below.
The coagulation equation can be discretized in terms of a monomer size dis-
tribution. In a monomer distribution, the volume of each particle in size bin k
equals the volume of a particle in the smallest size bin multiplied by k, as shown in
Fig. 15.1. In the monomer distribution, every possible particle volume that is a
multiple of the smallest volume is accounted for. Thus, coagulation always results
in particles moving to an exact bin instead of being partitioned between two bins.
For monomer size bins, the coagulation equation is discretized as
Ln
k
Lt
=
1
2
k1
j =1
kj. j
n
kj
n
j
n
k
j =1
k. j
n
j
(15.2)
where j and k are size-bin indices. Particles in bin k are produced when particles
in bin k j coagulate with particles in bin j. Bin k corresponds to particles with
volume
1
k, and bin j corresponds to particles with volume
1
j.
Equation (15.2) can be written in fully implicit nite-difference form (all vari-
ables on the right side are evaluated at time t) as
n
k.t
n
k.th
h
=
1
2
k1
j =1
P
k. j

j =1
L
k. j
(15.3)
where h is the time step, subscripts t and t h are the nal and initial times,
respectively, and
P
k. j
=
kj. j
n
kj.t
n
j.t
L
k. j
=
k. j
n
k.t
n
j.t
(15.4)
495
Coagulation
are production and loss rates, respectively. Note that P
k. j
= L
kj. j
. Substituting
(15.4) into (15.3) and rearranging gives
n
k.t
= n
k.th
+
1
2
h
k1
j =1
kj. j
n
kj.t
n
j.t
h
j =1
k. j
n
k.t
n
j.t
(15.5)
which represents a coagulation equation for one monomer size bin k.
15.2 SEMIIMPLICIT COAGULATION
Equation (15.5) can be solved with an iterative ordinary differential equation solver
(e.g., Jacobson and Turco 1995) or with one of several other techniques (e.g., Suck
and Brock 1979; Gelbard and Seinfeld 1980; Tsang and Brock 1982; Friedlander
1983; Strom et al. 1992; Fassi-Fihri et al. 1997; Trautmann and Wanner 1999;
Fernandez-Diaz et al. 2000; Sandu 2002). The scheme discussed here (Jacobson
et al. 1994; Toon et al. 1988; Turco et al. 1979) is called the semiimplicit coagulation
solution. It is named such because it is based on the substitution of n
j.th
for n
j.t
in
the production and loss terms of (15.4). The terms are now written in semiimplicit
form as
P
k. j
=
kj. j
n
kj.t
n
j.th
L
k. j
=
k. j
n
k.t
n
j.th
(15.6)
respectively. The use of these terms allows a noniterative solution to coagulation.
Combining (15.3) and (15.6) gives the semiimplicit solution for the number con-
centration of monomer particles after a time step as
n
k.t
= n
k.th
+
1
2
h
k1
j =1
kj. j
n
kj.t
n
j.th
h
j =1
k. j
n
k.t
n
j.th
(15.7)
which can be rewritten as
n
k.t
=
n
k.th
+
1
2
h
k1
j =1
kj. j
n
kj.t
n
j.th
1 +h
j =1
k. j
n
j.th
(15.8)
In this equation, k varies from 1 to any desired number of monomer size bins.
Although (15.8) correctly accounts for the reduction in particle number when two
particles coagulate (reducing the number by one-half), it does not conserve volume.
Equation (15.5) correctly accounts for number and volume, but it is fully implicit so
requires iteration to solve. In order to conserve volume and volume concentration
(which coagulation physically does) while giving up some accuracy in number con-
centration, (15.8) can be rederived in terms of volume concentrations of monomer
496
15.2 Semiimplicit coagulation
particles and rewritten as
v
k.t
=
v
k.th
+h
k1
j =1
kj. j
v
kj.t
n
j.th
1 +h
j =1
k. j
n
j.th
(15.9)
where v
k.t
=
k
n
k.t
from(13.10). Equation (15.9) satises the volume-conservation
requirement,
kj
P
k. j
=
kj
L
kj. j
, for each k and j.
The solution just presented is applicable to a monomer size distribution. It is
usually desirable to solve coagulation over an arbitrary size distribution, such as
the volume-ratio distribution. As discussed in Chapter 13, a volume-ratio size dis-
tribution is dened so that the volume of a particle in one size bin equals the volume
of a particle in the next smallest size bin multiplied by a constant, V
rat
.
For the volume ratio or any other type of discretized size distribution, the semi-
implicit solution can be modied to treat coagulation. The solution is found by
rst dening the volume of an intermediate particle that results when a particle in
size bin i collides and sticks to a particle in size bin j as
V
i. j
=
i
+
j
(15.10)
The intermediate particle has volume between those of two arbitrary model size
bins, k and k + 1, and needs to be partitioned between the two bins. This is done
by dening the volume fraction of V
i,j
that is partitioned to each model bin k as
f
i. j.k
=
k+1
V
i. j
k+1

k
V
i. j
k
V
i. j
-
k+1
k - N
B
1 f
i. j.k1

k1
- V
i. j
-
k
k > 1
1 V
i. j

k
k = N
B
0 all other cases
(15.11)
(Jacobson et al. 1994). The volume fractions in (15.11) are independent of the
size distribution. They work for a monomer distribution (where all values of f are
1 or 0), a volume-ratio distribution with any value of V
rat
> 1, or an arbitrary
distribution. Note that the volume fractions in (15.11) differ from the number
fractions used for partitioning number concentration with the quasistationary size
structure, given in (13.34).
When volume fractions are used, the semiimplicit, volume- and volume-
concentration conserving, noniterative, positive-denite for any time step coag-
ulation solution for total particle volume concentration becomes
v
k.t
=
v
k.th
+h
k
j =1
k1
i =1
f
i. j.k
i. j
v
i.t
n
j.th
1 +h
N
B
j =1
(1 f
k. j.k
)
k. j
n
j.th
(15.12)
497
Coagulation
The series of equations must be solved in the order k = 1, . . . , N
B
. No production
occurs into the rst bin, k = 1, since for it, k 1 = 0 in the numerator of (15.12).
Thus, all v
i.t
terms are known when v
k.t
is calculated. Values of f
i,j,k
in (15.12) are
frequently zero. During computer simulations, all such multiplications by zero are
eliminated.
The nal number concentration of particles due to coagulation is now simply
n
k.t
=
v
k.t
k
(15.13)
If particles have multiple components, the change in volume concentration of each
component q is calculated analogously to (15.12) with
v
q.k.t
=
v
q.k.th
+h
k
j =1
k1
i =1
f
i. j.k
i. j
v
q.i.t
n
j.th
1 +h
N
B
j =1
[(1 f
k. j.k
)
k. j
n
j.th
]
(15.14)
where the sum of v
q.k.t
over all components q = 1, . . . , N
V
is v
k.t
.
The advantage of a semiimplicit over a fully implicit equation is that the semi-
implicit equation yields a noniterative volume-conserving, volume concentration-
conserving stable solution to coagulation for any time step. Afully implicit equation
requires the use of an iterative ordinary-differential-equation solver. The disadvan-
tage of a semiimplicit equation is that particle number is not calculated exactly. By
increasing the resolution of the size distribution, the error in number concentration
approaches zero while the solution remains noniterative and volume- and volume
concentration-conserving.
15.3 COMPARISON WITH ANALYTICAL SOLUTIONS
In this section, results from the semiimplicit coagulation solution are compared
with analytical and numerical solutions. Smoluchowskis (1918) analytical solution
assumes that all particles are initially monodisperse (have the same size), and the
coagulation kernel is constant for all interactions. After initialization, coagulation
occurs over a monomer size distribution (Section 15.1). Atmospheric particles are
really polydisperse (particle number concentration varies with size). Nevertheless,
Smoluchowskis analytical solution is useful for checking the accuracy of numerical
coagulation schemes. Smoluchowskis solution is
n
k.t
=
n
T.th
(0.5hn
T.th
)
k1
(1 +0.5hn
T.th
)
k+1
(15.15)
where the subscripts t and t h refer to the nal and initial times, respectively,
n
T.th
is the initial number concentration of monodisperse particles in bin k = 1,
498
15.3 Comparison with analytical solutions
10
0
10
2
10
4
10
6
10
8 Initial
Smol.
Integrated
SI (1.2)
SI (1.5)
SI (2.0)
0.01 0.1
d
n

(
N
o
.

c
m
3
)

/
d

l
o
g
1
0

D
p
p
, m)
Figure 15.2 Comparison of Smoluchowskis ana-
lytical solution with an integrated numerical solu-
tion and with three semiimplicit solutions (SI), each
with a different value of V
rat
(given in parentheses).
The initial number concentration of monodisperse
particles was 10
6
particles cm
3
and T = 298 K.
This simulation time was 12 h. Remaining condi-
tions are described in the text.
n
k.t
is the nal concentration of particles in any monomer size bin k after time h,
the exponents k 1 and k + 1 refer to the size-bin number, and
=
8k
B
T
3
a
(15.16)
(cm
3
particle
1
s
1
) is the constant coagulation kernel. Equation (15.15) is solved
over one time step h of any size. Figure 15.2 compares Smoluchowskis solution
with an integrated numerical solution and three semiimplicit solutions to the same
problem over a full-stationary size structure with a volume-ratio size distribution.
The semiimplicit solutions improved with higher resolution (smaller V
rat
). The
integrated solution matched the analytical solution exactly, but the computer times
required for the semiimplicit solutions were much less than that required for the
integrated solution.
Another analytical solution is that for a self-preserving size distribution. Such a
distribution has initial number concentration in each size bin i of
n
i.th
=
n
T.th
L
i
p
exp
(15.17)
where n
T.th
is the initial total number concentration of particles among all size
bins,
i
is the volume of a single particle in size bin i,
p
is the volume at which
the initial peak number concentration occurs in the distribution, and L
i
is the
volume width of a size bin. For a volume-ratio size distribution, L
i
is calculated
with (13.8).
499
Coagulation
10
0
10
2
10
4
10
6
0.01 0.1 1
Initial
Analytical
SI (1.5)
SI (2)
d
n

(
N
o
.

c
m
3
)

/
d

l
o
g
1
0

D
p
p
, m)
Figure 15.3 Comparison of self-preserving
analytical solution with two semiimplicit
solutions (SI), each with a different value of
V
rat
(given in parentheses). The conditions
were n
T.th
= 10
5
cm
3
, T = 298 K, and
p
= 0.0005236 m
3
. The simulation time was
12 h.
After time h, the solution for coagulation over the self-preserving distribution is
n
i.t
=
n
T.th
L
i
p
(1 +0.5hn
T.th
)
2
exp
p
1 +0.5hn
T.th
(15.18)
where the coagulation kernel is given in (15.16). Figure 15.3 compares results from
the self-preserving analytical solution with those from two semiimplicit solutions.
15.4 COAGULATION AMONG MULTIPLE
PARTICLE DISTRIBUTIONS
The semiimplicit solution can be used to solve coagulation, not only over a single
size distribution with multiple components, but also among multiple size distribu-
tions, each with multiple components. Figure 15.4 illustrates coagulation interac-
tions among three externally mixed size distributions (A, B, and C, each represented
by one particle) that result in internally mixed particles of composition AB, AC,
A
B
C
AB = A+B = A+AB = B+AB
AC = A+C = A+AC = C+AC
BC = B+C = B+BC = C+BC
= A+B+C = AB+BC
= A+BC = AC+BC
= B+AC = AC+AB
ABC = C+AB = AB+ABC
= A+ABC = AC+ABC
= B+ABC = BC+ABC
= C+ABC = ABC+ABC
Figure 15.4 Example of internal mixing among three externally mixed size
distributions (A, B, and C, each represented by one particle). The gure shows
many (but not all) combinations resulting in internally mixed particles (AB,
AC, BC, and ABC).
500
15.4 Coagulation among multiple particle distributions
BC, or ABC. The gure and Example 15.1 show that many such combinations are
possible.
Example 15.1
What other possible combinations, not shown in Fig. 15.4, of three separate
particles can result in a particle of composition ABC?
SOLUTION
A+A+BC; B+B+AC; C+C+AB; A+AB+BC; B+AB+BC; C+AB+BC;
A+B+BC; A+B+AC; A+C+AB; A+C+BC; B+C+AB; B+C+AC.
Here, the semiimplicit coagulation solution is expanded to treat coagulation
among any number of size distributions (N
T
), size bins per distribution (N
B
),
and components per size bin per distribution (N
V
) (Jacobson 2002). The scheme
remains noniterative, positive-denite for any time step, and volume- and volume-
concentration-conserving.
With the scheme, the nal volume concentration (cm
3
cm
3
) of component q =
1, . . . , N
V
in size bin k = 1, . . . , N
B
of size distribution N = 1, . . . , N
T
at time t
is found by solving
v
q.Nk.t
=
v
q.Nk.th
+h(T
q.Nk.t.1
+ T
q.Nk.t.2
)
1 +hT
q.Nk.t.3
(15.19)
T
q.Nk.t.1
=
N
T
M=1
P
N.M
k
j =1
n
Mj.th
k1
i =1
f
Ni.Mj.Nk.th
Ni.Mj.th
v
q.Ni.t
T
q.Nk.t.2
=
N
T
M=1
N
T
I=1
Q
I.M.N
k
j =1
n
Mj.th
k
i =1
f
Ii.Mj.Nk.th
Ii.Mj.th
v
q.Ii.t
T
q.Nk.t.3
=
N
B
j =1
N
T
M=1
[(1 L
N.M
)(1 f
Nk.Mj.Nk.th
) + L
N.M
]
Nk.Mj.th
n
Mj.th
where term T
1
accounts for production of larger particles in distribution N from
self-coagulation and from heterocoagulation of distribution N with distribution
M = N. Term T
2
accounts for production of particles in distribution N from
heterocoagulation of two independent distributions (I = N and M= N). The rst
part of term T
3
accounts for self-coagulation loss in distribution N to larger sizes.
The second part accounts for loss of distribution N to all other distributions aside
from N due to heterocoagulation of N with the other distributions.
Equation (15.19) must be solved in a special order for all implicit (at time t) val-
ues on the right side of the equation to be known when they are used. Distributions
that have no coagulation production from other distributions are solved rst, fol-
lowed by distributions with production terms frompreviously solved distributions.
501
Coagulation
Take, for example, the set of 16 distributions shown in Table 15.1. This table lists
ve externally mixed distributions (A, B, D, E, and F), 10 binary distributions
(AB, AD, AE, AF, BD, BE, BF, DE, DF, and EF), and one well-mixed distribution
(MX = ABDEF). Due to the computational burden, ternary (e.g., ABD, ADE) and
quartenary (e.g., ABDE, ABDF) combinations are included in the MX distribu-
tion. In other words, coagulation between distributions AB and D, for example, is
assumed to produce mixed particles in distribution ABDEF.
In this example set of distributions, the externally mixed distributions are solved
rst (in any order), followed by the binary distributions (in any order), followed
by the MX distribution. Within each distribution, equations are solved from bin
k = 1, . . . , N
B
. The order of solving each volume component, q = 1, . . . , N
V
,
does not matter. To conserve volume at the upper boundary, (15.19) assumes that
particles cannot self-coagulate out of the largest size bin of any size distribution.
For example, when two particles self-coagulate to a size larger than the largest bin
in the distribution, they are placed in the largest bin but particle number is increased
to conserve volume concentration. Similarly, when two particles heterocoagulate
to form a new particle larger than the largest bin in the new distribution that the
particle enters, the new particle is placed in the largest bin of the new distribution
and number concentration is adjusted. These boundary conditions have virtually
no effect on results so long as the largest bin is much larger than that of the median
particle size in the distribution.
The parameters P, Q, and L in (15.19) are either 1 or 0, depending on the
coagulation interactions accounted for. The values are illustrated by considering
Table 15.1. The parameter P
N.M
= 1 if particles in distribution N coagulating with
particles in distribution M produce larger particles in distribution N. For example,
sea spraysoil (AB) plus sea spray (A) produces more sea spraysoil (AB); thus,
P
AB.A
= 1. The parameter Q
I.M.N
=1 if particles in distribution I coagulating with
particles in distribution M produce particles in distribution N, where I = M and
I = N. For example, soil (B) plus sulfate (D) produces soilsulfate (BD); thus,
Q
B.D.BD
= 1. The parameter L
N.M
= 1 if particles in distribution N coagulating
with particles in distribution M do not produce particles in distribution N. For
example, soil (B) plus sulfate (D) does not produce soil (B); thus, L
B.D
= 1. On
the other hand, soilsulfate (BD) plus sulfate (D) produces soilsulfate (BD); thus,
L
BD.D
= 0 in that case. Further, in (15.19)
f
Ii.Mj.Nk
=
Nk+1
V
Ii.Mj.th
Nk+1
Nk.th

Nk
V
Ii.Mj
Nk
V
Ii.Mj
-
Nk+1
k - N
B
1 f
Ii.Mj.Nk1

Nk1
- V
Ii.Mj
-
Nk
k > 1
1 V
Ii.Mj

Nk
k = N
B
0 all other cases
(15.20)
is the volume fraction of a coagulated pair, V
Ii.Mj
=
Ii
+
Mj
, that is partitioned
into bin k of distribution N. To minimize computer time, all calculations involving
a zero value of f, P, Q, or L are eliminated ahead of time.
502
T
a
b
l
e
1
5
.
1
C
o
a
g
u
l
a
t
i
o
n
i
n
t
e
r
a
c
t
i
o
n
s
f
o
r
t
h
e
s
i
m
u
l
a
t
i
o
n
i
n
F
i
g
.
1
5
.
5
S
y
m
b
o
l
o
f
s
e
c
o
n
d
d
i
s
t
r
i
b
u
t
i
o
n
S
i
z
e
d
i
s
t
r
i
b
u
t
i
o
n
n
a
m
e
/
s
y
m
b
o
l
A
B
D
E
F
A
B
A
D
A
E
A
F
B
D
B
E
B
F
D
E
D
F
E
F
M
X
S
e
a
s
p
r
a
y
A
A
A
B
A
D
A
E
A
F
A
B
A
D
A
E
A
F
M
X
M
X
M
X
M
X
M
X
M
X
M
X
S
o
i
l
B
A
B
B
B
D
B
E
B
F
A
B
M
X
M
X
M
X
B
D
B
E
B
F
M
X
M
X
M
X
M
X
S
u
l
f
a
t
e
D
A
D
B
D
D
D
E
D
F
M
X
A
D
M
X
M
X
B
D
M
X
M
X
D
E
D
F
M
X
M
X
E
m
i
t
t
e
d
s
o
o
t
(
E
S
)
E
A
E
B
E
D
E
E
E
F
M
X
M
X
A
E
M
X
M
X
B
E
M
X
D
E
M
X
E
F
M
X
B
a
c
k
g
r
o
u
n
d
s
o
o
t
(
B
S
)
F
A
F
B
F
D
F
E
F
F
M
X
M
X
M
X
A
F
M
X
M
X
B
F
M
X
D
F
E
F
M
X
S
e
a
s
p
r
a
y
s
o
i
l
A
B
A
B
A
B
M
X
M
X
M
X
A
B
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
S
e
a
s
p
r
a
y
s
u
l
f
a
t
e
A
D
A
D
M
X
A
D
M
X
M
X
M
X
A
D
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
S
e
a
s
p
r
a
y
E
S
A
E
A
E
M
X
M
X
A
E
M
X
M
X
M
X
A
E
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
S
e
a
s
p
r
a
y
B
S
A
F
A
F
M
X
M
X
M
X
A
F
M
X
M
X
M
X
A
F
M
X
M
X
M
X
M
X
M
X
M
X
M
X
S
o
i
l
s
u
l
f
a
t
e
B
D
M
X
B
D
B
D
M
X
M
X
M
X
M
X
M
X
M
X
B
D
M
X
M
X
M
X
M
X
M
X
M
X
S
o
i
l
E
S
B
E
M
X
B
E
M
X
B
E
M
X
M
X
M
X
M
X
M
X
M
X
B
E
M
X
M
X
M
X
M
X
M
X
S
o
i
l
B
S
B
F
M
X
B
F
M
X
M
X
B
F
M
X
M
X
M
X
M
X
M
X
M
X
B
F
M
X
M
X
M
X
M
X
S
u
l
f
a
t
e
E
S
D
E
M
X
M
X
D
E
D
E
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
D
E
M
X
M
X
M
X
S
u
l
f
a
t
e
B
S
D
F
M
X
M
X
D
F
M
X
D
F
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
D
F
M
X
M
X
E
S
B
S
E
F
M
X
M
X
M
X
E
F
E
F
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
E
F
M
X
I
n
t
e
r
n
a
l
m
i
x
t
u
r
e
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
M
X
T
h
e
t
a
b
l
e
g
i
v
e
s
t
h
e
s
y
m
b
o
l
o
f
t
h
e
s
i
z
e
d
i
s
t
r
i
b
u
t
i
o
n
i
n
t
o
w
h
i
c
h
t
w
o
p
a
r
t
i
c
l
e
s
f
r
o
m
e
i
t
h
e
r
t
h
e
s
a
m
e
o
r
d
i
f
f
e
r
e
n
t
d
i
s
t
r
i
b
u
t
i
o
n
s
c
o
a
g
u
l
a
t
e
.
F
o
r
e
x
a
m
p
l
e
,
w
h
e
n
a
s
e
a
s
p
r
a
y
p
a
r
t
i
c
l
e
(
A
)
c
o
a
g
u
l
a
t
e
s
w
i
t
h
a
s
o
i
l
p
a
r
t
i
c
l
e
(
B
)
,
t
h
e
r
e
s
u
l
t
i
n
g
p
a
r
t
i
c
l
e
i
s
a
s
e
a
s
p
r
a
y
s
o
i
l
(
A
B
)
p
a
r
t
i
c
l
e
.
W
h
e
n
t
h
r
e
e
o
r
m
o
r
e
e
x
t
e
r
n
a
l
l
y
m
i
x
e
d
d
i
s
t
r
i
b
u
t
i
o
n
s
(
A
,
.
.
.
,
F
)
o
r
a
n
e
x
t
e
r
n
a
l
l
y
m
i
x
e
d
a
n
d
b
i
n
a
r
y
d
i
s
t
r
i
b
u
t
i
o
n
(
A
B
,
.
.
.
,
E
F
)
o
r
t
w
o
b
i
n
a
r
y
d
i
s
t
r
i
b
u
t
i
o
n
s
c
o
m
b
i
n
e
,
t
h
e
y
e
n
t
e
r
t
h
e
m
i
x
e
d
d
i
s
t
r
i
b
u
t
i
o
n
(
M
X
)
.
T
h
u
s
,
t
h
e
c
o
m
b
i
n
a
t
i
o
n
o
f
b
a
c
k
g
r
o
u
n
d
s
o
o
t
(
F
)
w
i
t
h
s
e
a
s
p
r
a
y
s
o
i
l
(
A
B
)
p
r
o
d
u
c
e
s
a
m
i
x
e
d
p
a
r
t
i
c
l
e
(
M
X
)
.
Coagulation
10
6
10
4
10
2
10
0
10
2
10
4
10
6
0.01 0.1 1 10 100
A (Spray)
B (Soil)
D (Sulf)
E (ES)
F (BS)
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0

D
p
p
, m)
(a)
10
6
10
4
10
2
10
0
10
2
10
4
10
6
10
8
10
10
0.01 0.1 1 10 100
A (Spray)
B (Soil)
D (Sulf)
E (ES)
F (BS)
AB (Spray-soil)
AD (Spray-sulf)
AE (Spray-ES)
AF (Spray-BS)
BD (Soil-sulf)
BE (Soil-ES)
BF (Soil-BS)
DE (Sulf-ES)
DF (Sulf-BS)
EF (ES-BS)
MX
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0

D
p
p
, m)
(b)
10
6
10
4
10
2
10
0
10
2
10
4
0.01 0.1 1 10 100
Initial
12 hr
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0

D
p
p
, m)
(c)
Figure 15.5 (a) Initial number concentration of ve exter-
nally mixed size distributions. (b) Number concentrations
of 16 size distributions listed in Table 15.1 following 12
hours of coagulation starting with the ve distributions
in (a). (c) The initial and nal number concentrations,
summed over all distributions, from (a) and (b), respec-
tively. Following Jacobson (2002).
504
15.5 Particle flow regimes
The nal particle total volume concentration in each bin of each distribution is
v
Nk.t
=
N
V
q=1
v
q.Nk.t
(15.21)
Alternatively, the total volume concentration can be solved by summing the com-
ponent volume concentrations at time t h then solving (15.19) after replacing
v
q.Nk.t
with v
Nk.t
. Finally, the nal number concentration of particles in each bin of
each distribution is
n
Nk.t
=
v
Nk.t
Nk
(15.22)
Figure 15.5 illustrates the effect of coagulation among multiple size distribu-
tions on the internal mixing of aerosol particles. Coagulation alone was simulated
among the 16 distributions in Table 15.1 for a 12-hour period. Five of the distribu-
tions (A, B, D, E, F) were initialized, as shown in Fig. 15.5(a). Figure 15.5(b) shows
all 16 distributions after coagulation, and Fig. 15.5(c) shows the sum of all initial
and nal distributions. The last gure appears to indicate, as coagulation simu-
lations of one size distribution have indicated, that coagulation among multiple
size distributions has little effect on particles larger than 0.2 m in diameter over a
12-hour period. Figure 15.5(b) claries this misconception. It shows that, although
coagulation among multiple distributions has little effect on the total number con-
centration of large particles summed over all distributions, it internally mixes par-
ticles of different original compositions, affecting the composition of each size
distribution.
The results also suggest that, in an air mass containing a moderate loading of
particles, coagulation alone can internally mix most particles to some degree within
12 hours. In the simulation, coagulation internally mixed almost all the largest
particles and a smaller percentage of the smaller particles, which may explain a
result of Okada and Hitzenberger (2001), who found that the number fraction
of mixed particles in Vienna increased with increasing particle radius and with
increasing particle abundance.
Finally, the results suggest that coagulation may cause larger particles to become
mixed with more components than smaller particles. Since the combination of even
the smallest with the largest particle is considered a mixture, though, an internally
mixed large particle can often contain a trivial amount of a second and/or third
component.
15.5 PARTICLE FLOW REGIMES
For most of the rest of this chapter, physical processes causing particles to collide
and coalesce (coagulate) are discussed. The rate of particle collision during coag-
ulation depends on several parameters, including the Knudsen number for air, the
mean free path of an air molecule, the thermal speed of an air molecule, and the
particle Reynolds number. In this subsection, these parameters are described.
505
Coagulation
The Knudsen number for air, a dimensionless number, is dened as
Kn
a.i
=

a
r
i
(15.23)
where
a
is the mean free path of an air molecule, and r
i
is the radius of a particle
of size i. The mean free path of an air molecule is the average distance an air
molecule can travel before it encounters another air molecule by random motion
and exchanges momentum with it. It is analogous to the mixing length of an eddy,
dened in Chapter 8, and may be calculated as
a
=
2
a
a
v
a
=
2
a
v
a
(15.24)
where
a
is the dynamic viscosity of air (g cm
1
s
1
),
a
is the density of air (g cm
3
),
v
a
is thermal speed of an air molecule (cm s
1
), and
a
=
a
a
is the kinematic
viscosity of air, dened in (4.55). The greater the air density, the closer molecules
are to each other, and the shorter the mean free path of a molecule. The dynamic
viscosity of air was given empirically as a function of temperature in (4.54). The
average thermal speed of an air molecule, from (2.3), is
v
a
=
8k
B
T

M
(15.25)
The higher the air temperature, the greater the kinetic energy and average speed of
an air molecule.
Example 15.2
Find
a
and Kn
a.i
when T = 288 K, p
a
= 1013 hPa, the air is dry, and r
i
= 0.1
m.
SOLUTION
From(15.25), :
a
= 4.59 10
4
cms
1
. From(4.54),
a
= 1.79 10
4
g cm
1
s
1
.
Since the air is dry, R
m
= R
and
a
= 0.00123 g cm
3
from (2.36). Substitut-
ing these values into (15.24) gives
a
= 6.34 10
6
cm. From (15.23), Kn
a.i
=
0.63.
When a particle is large relative to the mean free path of an air molecule (Kn
a.i
1), the particle is likely to be intercepted by many air molecules, and its resistance
to motion is due primarily to air viscosity. This Knudsen-number regime is called
the continuum regime because particles see air as a continuum.
506
15.5 Particle flow regimes
10
11
10
7
10
3
10
1
10
5
0.01 0.1 1 10 100 1000
Knudsen
number
Diffusion coef. (cm
2
s
1
)
Reynolds
number
Figure 15.6 Knudsen number for air,
Reynolds number, and diffusion coefcient
of particles falling as a function of diameter
when T = 292 K, p
a
= 999 hPa, and
p
=
1.0 g cm
3
.
When the mean free path of an air molecule is large relative to particle size
(Kn
a.i
10), a particle is likely to be intercepted by relatively few air molecules,
and its resistance to motion is due primarily to the inertia of air molecules hitting it.
This Knudsen-number regime is called the free-molecular regime because particle
motion is governed by free-molecular kinetics. Between the continuum and free-
molecular regimes is the transition regime.
The particle Reynolds number gives the ratio of the inertial force exerted by a
particle to the viscous force exerted by air. If the inertial force is due to the particles
fall speed, the Reynolds number (dimensionless) is
Re
i
= 2r
i
V
f.i
a
(15.26)
where V
f,i
is the terminal fall speed of the particle (cm s
1
), calculated in Section
20.1. Figure 15.6 shows Reynolds and Knudsen numbers for particles falling under
specied conditions.
When particle radius is less than the mean free path of a gas molecule, the
particle motion through the air is called slip ow, because particles slip through
the air with little viscous resistance. Slip ow occurs mostly in the free-molecular
regime. Pure slip ow for falling particles occurs at low Reynolds numbers (e.g.,
Re
i
- 10
6
), since small, falling particles have a low terminal fall speed, as shown
in Fig. 15.6.
When the radius of a particle is larger than that of an air molecule, but the inertial
force imparted by the particle is negligible compared with the viscous force exerted
by air, the owof the particle through air is Stokes ow, named after George Stokes
(18191903). Stokes owoccurs within the continuumregime, since in Stokes ow
the particle radius is larger than the mean free path of an air molecule. Stokes ow
for falling particles occurs at moderate Reynolds numbers (e.g., 0.01-Re
i
-1). At
higher Reynolds numbers, the inertial force imparted by a particle becomes larger
than the viscous force exerted by air. At smaller Reynolds numbers, Stokes ow
approaches slip ow.
507
Coagulation
15.6 COAGULATION KERNEL
The coagulation rate coefcient (kernel),
i. j
= E
coal.i. j
K
i. j
(15.27)
(cm
3
particle
1
s
1
) is the product of a coalescence efciency (E
coal.i. j
, dimension-
less) and a collision kernel (K
i. j
, cm
3
particle
1
s
1
), where the subscripts indicate
particles in size bin i are coagulating with particles in size bin j. Equation (15.27)
assumes coagulation is occurring in only one size distribution, but the equation is
easily extended to multiple distributions.
For aerosolaerosol coagulation, the collision kernel is the sumof the kernels due
to the following physical processes: Brownian motion, convective Brownian motion
enhancement, gravitational collection, turbulent inertial motion, turbulent shear,
and van der Waals forces. The kernel is also affected by particle shape. For aerosol
hydrometeor coagulation and hydrometeorhydrometeor coagulation, the kernels
are also affected by thermophoresis, diffusiophoresis, and electric charge. Collision
kernels and coalescence efciencies for aerosolaerosol, aerosolhydrometeor, and
hydrometeorhydrometeor coagulation are discussed in this section. Numerical
solutions for solving aerosolhydrometeor and hydrometeor coagulation are given
in Chapter 18.
15.6.1 Brownian diffusion
Brownian motion is the irregular motion of a particle suspended in a gas due to
the randombombardment of the particle by gas molecules. Brownian motion leads
to particle scatter and is represented by a particle diffusion coefcient. As such, it
is often referred to as Brownian diffusion. However, particle diffusion is the net
transport of particles from a region of high concentration to one of low concen-
tration, whereas Brownian diffusion can occur in the absence of a concentration
gradient.
Brownian coagulation is the process by which particles scatter, collide, and
coalesce in the atmosphere due to Brownian motion. In the continuum regime, the
Brownian collision kernel (cm
3
particle
1
s
1
) for particles of size i coagulating
with particles of size j is
K
B
i. j
= 4(r
i
+r
j
)(D
p.i
+ D
p. j
) (15.28)
where D
p.i
and D
p. j
are the diffusion coefcients of particles i and j, respectively.
An expression for the particle diffusion coefcient is
D
p.i
=
k
B
T
6r
i
a
G
i
(15.29)
where k
B
is Boltzmanns constant, T is the Kelvin temperature, and G
i
= 1 +
Kn
a.i
a.i
is the Cunningham slip-ow correction (Cunningham 1910). When
508
15.6 Coagulation kernel
G
i
=1, (15.29) is called the StokesEinstein equation, which is valid for Stokes ow
only (0.01 - Re
i
- 1). When Re
i
- 0.01 (reaching slip ow at Re
i
- 10
6
), G
i
and
the diffusion coefcient increase because particle resistance to motion decreases,
allowing particles to move more freely through the air.
Knudsen and Weber (1911) approximated
a.i
= A
+ B
exp(C
Kn
a.i
), giving
the Cunningham slip-ow correction as
G
i
= 1 +Kn
a.i
[A
+B
exp(C
Kn
a.i
)] (15.30)
Knusden and Weber estimated A
, B
, and C
based on experiments with glass beads

falling through air. In a more relevant experiment, Millikan (1923) estimated A
,
B
, and C
as 0.864, 0.29, and 1.25, respectively, based on measurements of oil

drops falling through air. Kasten (1968) gives a more recent set of values for A
,
B
, and C
as 1.249, 0.42, and 0.87, respectively. Equation (15.30) predicts that,

at low Knudsen number (large particle radius and high Reynolds number), G
i
approaches unity, and the particle diffusion coefcient approaches that from the
StokesEinstein equation. At high Knudsen number (small particle radius and low
Reynolds number), G
i
increases, increasing the particle diffusion coefcient. Fig-
ure 15.6 shows the corrected particle diffusion coefcient as a function of particle
diameter. The gure shows that, at very low particle diameter (and Reynolds num-
ber), diffusion coefcients are slightly enhanced by the reduced resistance to particle
motion.
In the free-molecular regime, the Brownian collision kernel (cm
3
particle
1
s
1
)
for particles of size i coagulating with particles of size j is based on the kinetic
theory of gases and is given by
K
B
i. j
= (r
i
+r
j
)
2
v
2
p.i
+ v
2
p. j
(15.31)
where
v
p.i
=
8k
B
T

M
p.i
(15.32)
is the thermal speed of a particle in air, in which

M
p.i
is the mass (g) of one particle
of size i. Equation (15.32) indicates that the thermal speed of a light (small) particle
exceeds that of a heavy (large) particle at a given temperature.
For particles in the transition regime, the Brownian collision kernel can be cal-
culated with the interpolation formula of Fuchs (1964),
K
B
i. j
=
4(r
i
+r
j
)(D
p.i
+ D
p. j
)
r
i
+r
j
r
i
+r
j
+
2
i
+
2
j
+
4(D
p.i
+ D
p. j
)
v
2
p.i
+ v
2
p. j
(r
i
+r
j
)
(15.33)
where
i
is the mean distance from the center of a sphere reached by particles
leaving the spheres surface and traveling a distance of particle mean free path
p.i
.
509
Coagulation
These two parameters are, respectively,
i
=
(2r
i
+
p.i
)
3
4r
2
i
+
2
p.i
3
/
2
6r
i
p.i
2r
i

p.i
=
8D
p.i
v
p.i
(15.34)
Equation (15.33) simplies to (15.28) for small Knudsen numbers and to (15.31)
for large Knudsen numbers.
15.6.2 Convective Brownian diffusion enhancement
When a large particle falls through the air, eddies created in its wake enhance
diffusion of other particles to its surface. This process, called convective Brownian
diffusion enhancement, has the collision kernel
K
DE
i. j
=
K
B
i. j
0.45Re
13
j
Sc
13
p.i
Re
j
1. r
j
r
i
K
B
i. j
0.45Re
12
j
Sc
13
p.i
Re
j
> 1. r
j
r
i
(15.35)
(Pruppacher and Klett 1997), where Sc
p.i
is the particle Schmidt number,
Sc
p.i
=

a
D
p.i
(15.36)
which gives the ratio of viscous to diffusive forces. Equation (15.35) states that the
larger the particle and the higher its fall speed, the greater the effect of diffusion
enhancement.
15.6.3 Gravitational collection
Athird collision kernel is the gravitational collection (differential fall speed) kernel.
When two particles of different size fall, one may catch up with and collide with
the other. The collision kernel for gravitational collection accounts for this process
and is given as
K
GC
i. j
= E
coll.i. j
(r
i
+r
j
)
2
|V
f.i
V
f. j
| (15.37)
where E
coll.i. j
is a collision efciency. A parameterization of the collision efciency
of falling raindrops is
E
coll.i. j
=
60E
V.i. j
+ E
A.i. j
Re
j
60 +Re
j
r
j
r
i
(15.38)
E
V.i. j
=
1 +
0.75 ln(2St
i. j
)
St
i. j
1.214
2
St
i. j
> 1.214
0 St
i. j
1.214
E
A.i. j
=
St
2
i. j
(St
i. j
+0.5)
2
(15.39)
510
(Ludlum 1980) where St
i. j
= V
f.i
|V
f. j
V
f.i
|r
j
g (for r
j
r
i
) is the dimensionless
Stokes number. Equation (15.38) simplies to E
V.i. j
when Re
j
1 and to E
A.i. j
when Re
j
1. Gravitational collection is important where at least one particle is
large, such as in a cloud or in the presence of sea spray or soil dust particles. It
is the most important mechanism by which cloud drops form raindrops in warm
clouds (clouds with temperatures above 0

C).
Beard and Grover (1974) provide a set of measurements from a cloud-chamber
experiment of the collision kernel for liquid water drops resulting from Brownian
motion, Brownian diffusion enhancement, and gravitational collection simultane-
ously. Such measurements can be used instead of the model kernels provided above,
but are limited in the range of drop size interactions measured.
15.6.4 Turbulent inertial motion and turbulent shear
Two additional kernels include those for turbulent inertial motion and turbulent
shear. Turbulent inertial motion causes particles of different size to accelerate dif-
ferentially. The resulting collision kernel is
K
TI
i. j
=

34
d
g
14
a
(r
i
+r
j
)
2
|V
f.i
V
f. j
| (15.40)
(Saffman and Turner 1956), where
d
is the rate of dissipation of turbulent kinetic
energy per gram of medium (cm
2
s
3
), dened in Section 8.4. Values for
d
vary
between 3 and 2000 cm
2
s
3
, with lower values corresponding to clear or partly
cloudy air and larger values corresponding to strong cumulus convection. Atypical
value for clear air is 5 cm
2
s
3
When wind shear in turbulent air causes particles moving with the air to collide,
the collision kernel is parameterized with the kernel for turbulent shear,
K
TS
i. j
=
8
d
15
a
12
(r
i
+r
j
)
3
(15.41)
(Saffman and Turner 1956). Two small particles are more likely to collide because
of turbulent shear than because of turbulent inertial motion. A large particle is
more likely to collide with a small particle because of turbulent inertial motion
than because of turbulent shear. Brownian motions dominate both processes so
long as at least one particle is small.
Figures 15.7 (a) and (b) show how each of the ve coagulation kernels discussed
so far varies when particles of 0.01 and 10 m, respectively, coagulate with particles
of all sizes. The curves indicate that, for small particles, Brownian coagulation is
always dominant. For larger particles, the other kernels become important. When
two particles have the same size, their fall speeds are identical, and the coagulation
511
Coagulation
10
17
10
15
10
13
10
11
10
9
10
7
0.01 0.1 1 10
C
o
a
g
u
l
a
t
i
o
n

k
e
r
n
e
l

(
c
m
3

p
a
r
t
i
c
l
e
1

s
1
)
Radius of second particle (m)
Total
Brownian
Diff.
enhancement
Settling
Turb.
inertia
Turb.
shear
(a)
10
17
10
15
10
13
10
11
10
9
10
7
0.01 0.1 1 10
C
o
a
g
u
l
a
t
i
o
n

k
e
r
n
e
l

(
c
m
3

p
a
r
t
i
c
l
e
1

s
1
)
Radius of second particle (m)
Total
Brownian
Diff.
enhancement
Settling
Turb. inertia
Turb. shear
(b)
Figure 15.7 (a) Coagulation kernels for ve processes when a particle 0.01 m in
radius coagulates with particles of different size at 298 K. (b) Coagulation kernels
for ve processes when a particle 10 m in radius coagulates with particles of
different size at 298 K. The dip at 10 m results because the difference in fall speed
is zero at that point. The real width of the dip is narrower, but the resolution of the
size bins used for the graph was coarse.
kernels for turbulent inertial motions and gravitational collection are zero, as seen
in Fig. 15.7(b).
15.6.5 Van der Waals and viscous forces
Two forces that affect primarily small aerosol particles are van der Waals and
viscous forces. Van der Waals forces are weak dipoledipole attractions caused
by brief, local charge uctuations in nonpolar molecules having no net charge.
In other words, uncharged particles experience random charge uctuations that
cause one part of the particle to experience a brief positive charge and the other, a
brief negative charge, so that the particle still exhibits no net charge. When a par-
ticle experiencing a brief charge uctuation approaches another, the rst induces
a charge of the opposite sign on the closest end of the second particle. The oppo-
site charge between the two particles causes an attraction, enhancing the rate of
coagulation between the particles.
Viscous forces arise because two particles moving toward each other in a
viscous medium have a net diffusion coefcient smaller than that of the sum
of the two individual coefcients since velocity gradients generated by one
particle affect those of the other. Whereas van der Waals forces enhance the
rate of coagulation of small particles, particularly of nanoparticles (-50 nm
diameter), in the free-molecular regime, viscous forces retard the rate of van
der Waals force enhancement in the transition and continuum regimes (e.g.,
Marlow 1981; Schmidt-Ott and Burtscher 1982; Alam 1987; Seinfeld and Pandis
1998).
An interpolation formula for the van der Waals/viscous collision kernel between
the free-molecular and continuum regimes (Alam 1987; Jacobson and Seinfeld
512
2004) is
K
V
i. j
= K
B
i. j
V
E.i. j
1
= K
B
i. j
W
c.i. j
1 +
4(D
p.i
+ D
p. j
)
v
2
p.i
+ v
2
p. j
(r
i
+r
j
)
1 +
W
c.i. j
W
k.i. j
4(D
p.i
+ D
p. j
)
v
2
p.i
+ v
2
p. j
(r
i
+r
j
)
1
(15.42)
where V
E.i. j
is the Van der Waals/viscous collision correction factor (which may
be less than or greater than unity). Although K
V
i. j
may be less than zero, K
V
i. j
+
K
B
i. j
> 0.
The individual correction factors in (15.42) for the free-molecular and contin-
uum regimes are
W
k.i. j
=
1
2(r
i
+r
j
)
2
k
B
T

r
i
+r
j
dE
P.i. j
(r)
dr
+r
d
2
E
P.i. j
(r)
dr
2
exp
1
k
B
T
r
2
dE
P.i. j
(r)
dr
+ E
P.i. j
(r)
r
2
dr
(15.43)
W
c.i. j
=
1
(r
i
+r
j
)

r
i
+r
j
D
i. j
D
r.i. j
(r) exp
E
P.i. j
(r)
k
B
T
dr
r
2
(15.44)
respectively. In these equations,
E
P.i. j
(r) =
A
H
6
2r
i
r
j
r
2
(r
i
+r
j
)
2
+
2r
i
r
j
r
2
(r
i
r
j
)
2
+ln
r
2
(r
i
+r
j
)
2
r
2
(r
i
r
j
)
2
(15.45)
is the van der Waals interaction potential, in which A
H
is the Hamaker constant,
which depends on van der Waals properties of each substance (e.g., Seinfeld and
Pandis 1998). Values of A
H
k
B
T generally range from -20 to 200. Also,
D
i. j
D
r.i. j
(r) = 1 +
2.6r
i
r
j
(r
i
+r
j
)
2
r
i
r
j
(r
i
+r
j
)(r r
i
r
j
)
+
r
i
r
j
(r
i
+r
j
)(r r
i
r
j
)
(15.46)
is a viscous-force correction factor to the diffusion coefcient in the continuum
regime. In the equation, D
r.i. j
is a relative diffusion coefcient between particles
i and j, and D
i. j
= D
p.i
+ D
p. j
is the sum of the individual diffusion coefcients of
the two particles.
513
Coagulation
0
1
2
3
4
5
6
0.01 0.1 1 10 100 1000
1
2
5
10
50
250
C
o
r
r
e
c
t
i
o
n

f
a
c
t
o
r
Particle Knudsen number
r
2
/r
1
Figure 15.8 Van der Waals/viscous force correction fac-
tor (V
E.i. j
) as a function of particle Knudsen number
(Kn
p
) and different ratios of the radii of two interacting
particles when T =300 K, p =1 atm, particle density =
1 g/cm
3
, and A
H
k
B
T = 200. From Jacobson and Sein-
feld (2004).
Figure 15.8 shows the van der Waals/viscous force correction factor V
E.i. j
versus
particle pair Knudsen number,
Kn
p
=
2
p.i
+
2
p. j
r
i
+r
j
(15.47)
assuming A
H
k
B
T =200. The gure shows that van der Waals forces enhanced the
coagulation rate of small particles (high Kn
p
) by up to a factor of ve. The calcu-
lated enhancement factor for two 14-nm diameter particles (Kn
p
= 2.26) from the
gure is about 3.25, and that for two 28-nm particles (Kn
p
= 0.983) is about 2.3.
These numbers compare with measured enhancement factors for 14- and 28-nm
graphitic carbon particles of 68 and 1.43, respectively (Burtscher and Schmidt-
Ott 1982). Although a value of A
H
k
B
T = 200 is higher than typical measured
values for many chemicals (e.g., Visser 1972), the comparison above illustrates that
A
H
k
B
T = 200 results in conservative enhancement factors for graphite. Similarly,
silver particles have measured enhancement factors much higher than those pre-
dicted using their Hamaker constant (Burtscher and Schmidt-Ott 1982; Okuyama
et al. 1984).
15.6.6 Fractal geometry
Fractals are particles of irregular, fragmented shape. Many aerosol particles con-
taining liquids, such as water, are roughly spherical. Soot aggregates, though, are
fractal in nature. Treatment of aggregates as fractal particles increases the rate of
coagulation (e.g., Mountain et al. 1986; Mulholland et al. 1988; Rogak and Flagan
1992; Harris and Maricq 2001; Artelt et al. 2003).
The effect of fractal geometry on collisions can be treated by considering
the effect of shape on radius, the diffusion coefcient, and the Knudsen num-
ber in previous collision kernels. For example, when an amorphous-shaped soot
514
particle consisting of an agglomerate of individual spherules is considered, the
volume-equivalent radius (r
i
, the radius of a sphere with the same volume and
density as the agglomerate) is replaced in the Brownian collision kernel of (15.33)
by different terms in different parts of the equation.
For example, some terms are replaced with the fractal radius (outer radius) of
the agglomerate,
r
f.i
= r
s
N
1D
f
i
(15.48)
In this equation, r
s
is the radius (cm) of each individual spherule (each assumed to
be the same size), D
f
is the fractal dimension, which relates the outer volume of an
aggregate to the summed volume of individual spherules, and
N
s.i
=

i
s
(15.49)
is the number of individual spherules in the soot aggregate, where
s
= 4r
3
s
3
is the volume of an individual spherule and
i
= 4r
3
i
3 is the volume of the
aggregate as if it were a sphere of uniform density. For diesel soot, individual
spherules have typical radii of about r
s
1015 nm (e.g., Weingartner et al. 1997;
Naumann 2003; Schnaiter et al. 2003; Wentzel et al. 2003). For pure spheres,
D
f
= 3. For diesel soot aggregates measured in the laboratory and in ambient
air, D
f
ranges from 1.5 to 2.5, with a typical value near 1.7 (e.g., Klingen and
Roth 1989; Katrlnak et al. 1993; Lee 2001; Xiong and Friedlander 2001; Wentzel
et al. 2003). As material condenses upon a soot aggregate, the material tends to
compress the aggregate closer to a spherical shape, increasing D
f
(Schnaiter et al.
2003; Wentzel et al. 2003).
Another type of radius used to calculate the effect of fractal geometry is the
mobility radius, which is the radius of a sphere moving at the same speed and
experiencing the same drag force as the agglomerate. This differs from the Stokes
radius, which is the radius of a sphere with the same density and moving at the
same fall speed as the agglomerate. It also differs from the aerodynamic radius,
which is the radius of a unit-density sphere that has the same fall speed as the
agglomerate. An expression for the mobility radius is
r
m.i
= MAX
r
f.i
ln(r
f.i
r
s
) +1
. r
f.i
D
f
1
2
0.7
. r
A.i
(15.50)
(Rogak and Flagan 1992) where
r
A.i
= r
s
MAX
N
23
s.i
. MIN
1 +
2
3
(N
s.i
1).
1
3
D
f
N
2D
f
s.i
(15.51)
is the area-equivalent radius (cm) (the radius of a sphere with the same surface area
as the agglomerate).
For determining collision rates among fractals, an addition radius, the collision
radius (r
c.i
), needs to be considered. Rogak and Flagan (1992) suggest that this
radius should lie between the fractal radius and the mobility radius. They also
515
Coagulation
found that their modeled coagulation kernel matched measurements the best when
the fractal (outer) radius was used as the collision radius (r
c.i
= r
f.i
), which is the
assumption used here. For spheres (D
f
= 3), the fractal radius, mobility radius,
area-equivalent radius, and collision radius are identical and equal the volume-
equivalent radius.
With the denitions above, the Brownian collision kernel (cm
3
particle
1
s
1
)
from (15.33) can be modied for fractal geometry with
K
B
i. j
=
4(r
c.i
+r
c. j
)(D
m.i
+ D
m. j
)
r
c.i
+r
c. j
r
c.i
+r
c. j
+
2
m.i
+
2
m. j
+
4(D
m.i
+ D
m. j
)
v
2
p.i
+ v
2
p. j
(r
c.i
+r
c. j
)
(15.52)
(Jacobson and Seinfeld 2004) where D
m
is a particle diffusion coefcient (cm
2
s
1
)
from (15.29) except that it and the Knudsen number for air, used within it, are
evaluated at the mobility radius, and
m
is the mean distance (cm) from (15.34),
except that it and the mean free path of a particle, used within it, are evaluated
at the mobility radius (Rogak and Flagan 1992). The mean thermal speed of a
particle ( v
p
), used in (15.52) and (15.34), is unaffected by particle radius. Instead,
it is affected by particle mass, which is independent of fractal dimension. Equation
(15.52) simplies to the Brownian kernel for spheres when D
f
= 3.
Figure 15.9 shows the effect of van der Waals/viscous forces and fractal geome-
try on the Brownian collision kernel. Van der Waals forces enhanced the kernel less
for fractals than for spherical particles because, for the same volume-equivalent
diameter, a fractal is more likely than a spherical particle to be in the transition
or continuum regime than the free-molecular regime, and van der Waals forces
are weakest in the continuum regime. The gure also shows that, when one par-
ticle was small (10 nm), fractal geometry increased the kernel with increasing size
of the second particle. When one particle was larger (100 nm), fractal geometry
enhancement was minimized when the second particle was also 100 nm.
Figure 15.10(a) shows the evolution, due to coagulation alone, of a size distri-
bution dominated by nanoparticles emitted near a freeway. The coagulation kernel
accounted for Brownian motion, Brownian diffusion enhancement, gravitational
collection, turbulent inertial motion, and turbulent shear but not van der Waals
forces, viscous forces, or fractal geometry. Of the kernels treated, Brownian motion
was by far the most important, causing over 97 percent of particle loss. The gure
shows that the rst peak disappeared in favor of the second peak of 40 nm after
3045 minutes.
Figure 15.10(b) shows the effect of van der Waals/viscous forces and fractal
geometry on the size distribution of aerosol particles simulated in Fig. 15.10(a).
The gure shows that treating these additional processes decreased the time of
removal of the rst peak from 3045 minutes to about 10 minutes. Thus, van der
Waals/viscous forces and fractal geometry may account for a greater share of the
evolution of nanoparticle size than Brownian motion alone (Jacobson and Seinfeld
2004).
516
10
9
10
8
10
7
10
6
10
5
10
4
0.01 0.1 1
Spherical, no van der Waals
Spherical, with van der Waals
Fractal, no van der Waals
Fractal, with van der Waals
C
o
l
l
i
s
i
o
n

k
e
r
n
e
l
(
c
m
3

p
a
r
t
i
c
.
1

s
1
)
Volume-equivalent diameter of the second particle (m)
(a)
10
9
10
8
10
7
0.01 0.1 1
Spherical, no van der Waals
Spherical, with van der Waals
Fractal, no van der Waals
Fractal, with van der Waals
C
o
l
l
i
s
i
o
n

k
e
r
n
e
l

(
c
m
3

p
a
r
t
i
c
.
1

s
1
)
Volume-equivalent diameter of the second particle (m)
(b)
Figure 15.9 Brownian collision kernel when the volume-
equivalent diameter of the rst particle is (a) 10 nm and (b)
100 nm and the volume-equivalent diameter of the second
particle varies from 10 to 1000 nm. The four curves shown
in each gure account for when particles are spherical (D
f
=3)
or fractal (D
f
= 1.7 above 27 nm diameter and = 3.0 below)
and when van der Waals and viscous forces are or are not
included. The collision radius was set to the fractal radius.
Also, A
H
k
B
T = 200, T = 300 K, p = 1 atm, and r
s
= 13.5
nm.
15.6.7 Diffusiophoresis, thermophoresis, and electric charge
Aerosolhydrometeor coagulation and hydrometeorhydrometeor coagulation are
affected by three additional processes not so important during aerosolaerosol
coagulation: diffusiophoresis, thermophoresis, and electric charge.
Diffusiophoresis is the ow of aerosol particles down a concentration gradient
of a gas due to the bombardment of the particles by the gas as the gas diffuses down
the same gradient. For example, when water vapors partial pressure exceeds its
saturation vapor pressure over a liquid cloud drop surface, water vapor ows to
the drop surface, enhancing the ow of aerosol particles in its path to the surface
as well. The increased ux of particles to the drop increases the rate of collision by
diffusiophoresis between the particles and the liquid drop. Evaporation of a drop
causes water vapor and particles to ow away from the drop surface, decreasing
the aerosolhydrometeor collision rate.
Thermophoresis is the ow of aerosol particles from warm air to cool air due to
bombardment of the particles by gases in the presence of a temperature gradient.
517
Coagulation
0 10
0
5 10
4
1 10
5
1.5 10
5
2 10
5
2.5 10
5
3 10
5
3.5 10
5
4 10
5
0.01 0.1 1
8 s
1 m
2 m
3 m
5 m
10 m
15 m
20 m
30 m
45 m
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0

D
p
p
, m)
(a)
0 10
0
1 10
5
2 10
5
3 10
5
4 10
5
0.01 0.1 1
8 s
1 m
2 m
3 m
5 m
10 m
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
D
1
0

p
p
, m)
(b)
Figure 15.10 Box-model calculation of the evolution of a size
distribution by coagulation when (a) particles are assumed
spherical and no van der Waals or viscous forces are acting
and (b) particles are assumed fractal (D
f
= 1.7 above 27 nm
diameter and = 3.0 below) and van der Waals and viscous
forces are acting ( A
H
k
B
T = 200). Other conditions were
T = 300 K and p = 1 atm. No dilution was accounted for. m
is minutes and s is seconds. The results were obtained by treat-
ing coagulation among 10 size distributions and summing the
results among all. From Jacobson and Seinfeld (2004).
When a temperature gradient exists, one side of an aerosol particle is warmer than
the other side. On the warm side, gas molecules have a greater thermal speed, thus
impart a greater force than they do on the cool side, pushing the particle fromwarm
to cool air. Near a cloud-drop surface, release of latent heat due to condensation of
water vapor causes the surface to be warmer than the air around it, enhancing ow
of aerosol particles away from the surface, decreasing the collision rate of aerosol
particles with the surface. Evaporation has the opposite effect, increasing the rate
of aerosolhydrometeor collisions due to thermophoresis.
The rate of coagulation of aerosol particles with cloud drops is also affected by
electric charge. Ice crystals, cloud drops, and aerosol particles are often charged.
Ice crystals, in particular, become charged when they collide with but bounce off
other ice crystals. The bounceoff often causes a charge separation, especially if a
temperature difference exists between colliding particles. Alternatively, a charged
ice crystal may induce a charge on a nearby aerosol particle or a cloud drop due
to electrostatic forces. Similarly, a charged aerosol particle may induce a charge
518
upon a cloud drop. Electrostatic forces, though, are important only when charge
differences are large, such as between an aerosol particle and a cloud drop or two
cloud drops, but not, for example, between two submicrometer aerosol particles
(e.g., Seinfeld and Pandis 1998).
A collision kernel for aerosol particles with cloud drops that accounts for diffu-
siophoresis, thermophoresis, electric charge, Brownian motion, Brownian diffusion
enhancement, turbulent shear, and turbulent inertial motion simultaneously is
K
i. j
=
4B
P.i
C
i. j
exp
4B
P.i
C
i. j
K
B
i. j
+ K
DE
i. j
+ K
TI
i. j
+ K
TS
i. j
1
(15.53)
(Wang et al. 1978; Martin et al. 1980 with addition of the turbulence terms by
Jacobson 2003). In this equation, B
P.i
is the mobility (s g
1
) of an aerosol particle,
which is the particles drift speed per unit external force acting on it. The drift speed
is the mean speed of a population of particles arising from the external force. For
example, if the only external force acting on an aerosol particle of spherical radius
r
i
(cm) falling through the air is the force of gravity, the drift speed of the particle
is its terminal fall speed, V
f.i
(Section 20.1) (cm s
1
), and the resulting mobility of
the particle is its terminal fall speed divided by the force of gravity,
B
P.i
=
V
f.i
F
G
=
V
f.i
F
D
=
G
i
6
a
r
i
=
D
p.i
k
B
T
(15.54)
In this equation,
F
G
=

M
p.i
g (15.55)
is the force of gravity (g cm s
2
), where

M
p.i
is single-particle mass (g), and
F
D
=
6r
i
a
V
f.i
G
i
(15.56)
(g cm s
2
) is the force of drag, which is equal in magnitude to but opposite in
direction from the force of gravity (Section 20.1). In addition, G
i
is the dimension-
less Cunningham slip-ow correction from (15.30),
a
is the dynamic viscosity of
air from (4.54) converted from units of kg m
1
s
1
to g cm
1
s
1
, and D
p.i
is the
particle diffusion coefcient from (15.29) (cm
2
s
1
). From (15.54), the diffusion
coefcient of the particle can be rewritten as
D
p.i
= B
P.i
k
B
T (15.57)
which is the Einstein relation.
The second new term in (15.53) is
C
i. j
= C
Th
i. j
+C
Df
i. j
+C
e
i. j
(15.58)
(g cm
3
s
2
), which is a parameter that is a function of the thermophoretic, diffusio-
phoretic, and electric forces acting on an aerosol particle in the presence of water
519
Coagulation
vapor and a cloud drop. The individual terms in the parameter are
C
Th
i. j
=
12r
i
a
(
a
+2.5
p
Kn
a.i
)
a
r
j
(T
T
s. j
)F
h.L. j
5(1 +3Kn
a.i
)(
p
+2
a
+5
p
Kn
a.i
) p
a
(15.59)
C
Df
i. j
= 6
a
r
i
0.74D
v
m
d
r
j
(
v
v.s
)F
v.L. j
G
i
m
v
a
(15.60)
C
e
i. j
= Q
i
Q
j
(15.61)
where r
i
and r
j
are the aerosol particle and cloud drop radii, respectively (cm),
Kn
a.i
is the dimensionless Knudsen number for air,
a
is the thermal conductivity
of moist air from (2.7) converted to units of J cm
1
s
1
K
1
,
p
0.00419 J cm
1
s
1
K
1
is the thermal conductivity of the aerosol particle, T
is the air temperature

far from the cloud drop (K), T
s. j
is the cloud-drop surface temperature (K), p
a
is
the air pressure (J cm
3
),
v
is the mass density of water vapor away from the drop
surface (g cm
3
),
v.s
is the saturation mass density of water over the surface from
Section 2.5.3 (g cm
3
),
a
3
), F
h.L. j
is the dimensionless thermal
ventilation factor from (16.31), F
v.L. j
is the ventilation coefcient for water vapor
from(16.24), D
v
is the diffusion coefcient of water vapor through air from(16.17)
(cm
2
s
1
), and m
d
and m
v
are the molecular weights of dry air and water vapor,
respectively (g mol
1
).
In addition, Q
i
= q
i
r
2
i
and Q
j
= q
j
r
2
j
are the electric charge on the aerosol
particle and water drop, respectively (esu, where 1 esu = 1 g
12
cm
32
s
1
= 3.33
10
10
C). In these terms, q
i
and q
j
are the charge per square centimeter of surface
(esu cm
2
). The theoretical maximum charge a particle can carry is the Rayleigh
limit for disruption,
Q
j.max
=
16
p
r
3
j
(15.62)
(esu) where
p
is the surface tension of the particle (dyn cm
1
=g s
2
). An empirical
t to the measured charge on cloud drops in highly electried, precipitating clouds
is
Q
j
= 2r
2
j
(15.63)
(esu), where r
j
is in cm. This equation gives q
j
= 2 esu cm
2
. An empirical t to
measured charges in warm clouds, which have weak electrication, is
Q
j
= 0.0005r
1.3
j
(15.64)
C
i. j
is valid for 0.001 - r
i
- 10 m and 42 - r
j
- 310 m. If gravitational
forces, charge, thermophoresis, and diffusiophoresis are ignored, (15.53) simplies
to K
i. j
= K
B
i. j
+ K
DE
i. j
+ K
TI
i. j
+ K
TS
i. j
.
520
10
5
10
4
10
3
10
2
10
1
10
0
0.001 0.01 0.1 1 10
C
o
l
l
i
s
i
o
n

e
f
f
i
c
i
e
n
c
y
Radius of aerosol particle (m)
e
d
=0, q=0
e
d
=0, q=2
e
d
=100, q=2
e
d
=100,q=0
Figure 15.11 Collision efciency from(15.53)
of a 42-m radius cloud drop (large par-
ticle) interacting with smaller aerosol parti-
cles for two dissipation rates (
d
, cm
2
s
3
)
and two charge conditions (q, esu cm
2
, set
to the same value for both particles). Other
conditions were T = 283.15 K, p
a
= 900
hPa, and RH = 75%. The collision efciency
equaled the total coagulation kernel divided by
(r
i
+r
j
)
2
|V
f. j
V
f.i
|. Eight collision kernels
were accounted for (see text). (Compare
d
=0
cases with Curves 2 and 6 of Fig. 4a of Wang
et al. 1978.)
Figure 15.11 shows modeled collision efciencies from (15.53), accounting for
Brownian motion, Brownian diffusion enhancement, gravitational collection, tur-
bulent shear, turbulent inertial motion, diffusiophoresis, thermophoresis, and elec-
tric charge for a 42-mradius cloud drop interacting with smaller aerosol particles
for two turbulent dissipation rates (
d
, cm
2
s
3
) and two charge conditions (q, esu
cm
2
). The addition of turbulence affected the collision efciency the greatest when
electric charge was zero. For nonzero or zero charge, turbulence increased collision
efciencies more for large particles than for small particles.
Figure 15.12 shows individual collision kernels when
d
= 100 cm
2
s
3
, q =
2 esu cm
2
(electried thunderstorm cloud) and other conditions were the same
as those in Fig. 15.11. Brownian motion and diffusion enhancement dominated
collision with a 42-m cloud drop when the aerosol-particle radius was -0.01
m. Turbulent inertia dominated when the aerosol-particle radius was 0.01
4 m. Gravitational collection dominated when the aerosol-particle radius was
>4 m.
15.6.8 Coalescence efciencies
When two aerosol particles collide, they may or may not stick together, depending
on the efciency of coalescence. The efciency of coalescence depends on par-
ticle shape, composition, ambient relative humidity, and other factors. Because
the kinetic energy of collision between one small particle (-2-m-radius) and
a small or large particle is relatively small in comparison with that between
521
Coagulation
10
10
10
9
10
8
10
7
10
6
10
5
10
4
10
3
10
2
0.001 0.01 0.1 1 10
C
o
l
l
i
s
i
o
n

k
e
r
n
e
l

(
c
m
3

p
a
r
t
i
c
.
1

s
1
)
Radius of small particle (m)
Brownian
Br. Dif. Enhanc.
Total
Turb. shear
Turb. Inert.
Grav.
Figure 15.12 Individual and total collision ker-
nels for the q = 2 esu cm
2
and
d
= 100 cm
2
s
3
case in Fig. 15.11.
two large particles, the likelihood that bounceoff occurs when a small particle
collides with another particle is low, and the coalescence efciency is approxi-
mately unity (Beard and Ochs 1984; Pruppacher and Klett 1997). To the con-
trary, van der Waals forces enhance coalescence between two small particles.
This enhancement is quantied by van der Waals/viscous force correction fac-
tor, V
E.i. j
. Thus, van der Waals/viscous forces can be treated either in terms
of a collision kernel (15.42) or in terms of a coalescence efciency (E
coal.i. j
=
V
E.i. j
).
When two colliding particles are large, the kinetic energy of collision is high,
increasing the chance of a bounceoff following collision, and the coalescence ef-
ciency is less than 1. Some parameterizations of coalescence efciency between
large particles, derived from cloud-chamber experiments, are given in Beard and
Ochs (1984) and Pruppacher and Klett (1997).
15.7 SUMMARY
The process of coagulation is described by the integrodifferential coagulation equa-
tion. This equation can be solved exactly with analytical solutions for a few ideal-
ized cases and with iterative numerical techniques for all cases. It can also be solved
approximately with a semiimplicit solution technique that conserves volume and
volume concentration and is noniterative and positive-denite. The technique also
allows simulation of coagulation among any number of particle size distributions
and among particles with multiple components. The rate of coagulation depends on
Brownian diffusion, enhancement of Brownian motion due to eddies, differences in
fall speeds, turbulent shear, turbulent inertial motion, van der Waals forces, viscous
forces, particle shape, diffusiophoresis, thermophoresis, and electric charge. Coag-
ulation kernels were described for each of these processes. Effects of coagulation
coupled to growth are discussed in Chapter 16. Coagulation in clouds is discussed
in more detail in Chapter 18.
522
15.8 PROBLEMS
15.1 If an air mass contains six homogeneous, externally-mixed particle types,
how many possible combinations of internal mixtures of 2, 3, 4, 5, and 6
particle types exist? What is the total number of externally-plus internally-
mixed particle types possible?
15.2 Suppose two particles of diameter 100 nm and 200 nm coagulate with each
other. (a) Calculate the volume of the resulting coagulated pair. (b) If a model
has xed size bins with mean diameters 200 nm and 300 nm, calculate the
volume fraction of the coagulated pair going into each xed bin. (c) Calculate
the number fraction of particles going into each xed size bin.
15.3 Use Smoluchowskis solution to calculate the number concentration of par-
ticles with volume 10 times that of the initial monodisperse distribution
after 12 hours of coagulation when the temperature is 298 K assuming the
initial concentration of monodisperse particles is 10
6
particles cm
3
.
15.4 Calculate the coagulation kernel due to Brownian motion when the radius
of one particle is 0.1 m and the radius of the other is 1.0 m. Compare the
result with the kernel obtained when both particles are 1.0 m in radius.
Assume T = 298 K, p
d
= 990 hPa, the air is dry, and the particle density is
p
= 1.2 g cm
3
.
15.5 Calculate the turbulent inertial motion coagulation kernel between 0.5-m
and 10-m radius liquid water drops assuming a dissipation rate for tur-
bulent kinetic energy of (a) 5 cm
2
s
3
and (b) 2000 cm
2
s
3
. Assume the
temperature is 298 K, the air density is 0.001 225 g cm
3
, and the fall speed
can be approximated by V
f.i
2r
2
i

p
9
a
, where
p
=drop density (1 g cm
3
).
15.6 If the volume-equivalent radius of a soot agglomerate with spherule radius
of 27 nm is 100 nm, and the agglomerates fractal dimension is 1.7, calculate
its (a) fractal radius, (b) area-equivalent radius, and (c) mobility radius.
15.7 Give two reasons why the coalescence efciency of two nanoparticles is near
unity.
15.8 Using the constant coagulation kernel from (15.16), plot Smoluchowskis
analytical solution on a dndlog
10
D
p
graph when T = 298 K and h = 12 h
(dn = n
j
and dlog
10
D
p
= log
10
d
i+1
log
10
d
i j
for each i). Assume that the
initial total particle number concentrationis 10
6
particles cm
3
andall initial
particles have diameter 0.006 m.
15.9 Write a computer script to calculate the coagulation kernel due to Brownian
motion as a function of temperature and particle size. Use a volume ratio
size distribution, with d
1
= 0.001 m, d
N
B
= 10 m, and V
rat
= 1.5. Assume
p
d
= 1013 hPa, the air is dry, and the particle mass density is
p
= 1.5 g
cm
3
. Calculate the coagulation kernels when (a) T = 298 K, the diameter
of one particle (d
a
) is 0.001 m, and the diameter of the other particle (d
b
)
varies from 0.001 to 10 m; (b) T = 298 K, d
a
= 10 m, and d
b
varies from
0.001 to 10 m; (c) the diameters of both particles are 0.001 m and T varies
from 190 to 320 K in increments of 5 K; and (d) d
a
= 0.001 m, d
b
= 10 m,
523
Coagulation
and T varies from 190 to 320 K in increments of 5 K. Plot and interpret the
results.
15.10 Write a computer programto simulate semiimplicit coagulationwith(15.12).
Use the program with the coagulation kernel and conditions described in
Problem15.8 to simulate Smoluchowskis solution. Compare the results with
those from Problem 15.8.
15.11 Write a computer program to calculate the collision kernel for a 0.5-m
aerosol particle with a 10-mradius liquid water drop using (15.53), neglect-
ing turbulent shear and turbulent inertial motion. Assume the temperature
away from cloud drop surfaces is 298 K and that at drop surfaces is 300 K.
Also assume that the air density is 0.001 225 g cm
3
, the relative humidity
away from cloud drop surfaces is 90 percent, and the fall speed is approx-
imately V
f.i
2r
2
i

p
9
a
, where
p
= drop density (1 g cm
3
for both the
aerosol particle and cloud drop). Ignore ventilation coefcients and electric
charge. Once the code has been tested for one pair of particles, plot the kernel
for coagulation between the 10-m cloud drop and aerosol particles varying
from 0.001 to 10m in radius.
524
16
Condensation, evaporation, deposition,
and sublimation
A
parti cle grows in size by the mass transfer of gas to its surface followed
by the change in state or chemical conversion of the gas at the surface by
condensation, solid deposition, chemical reaction, and/or dissolution. In this chap-
ter, equations for vapor condensation and ice deposition on particle surfaces are
discussed. Equations for water-vapor condensation onto a single homogeneous
particle are rst derived. The growth equations are extended to gases other than
water vapor, to a population of particles, and to particles with multiple compo-
nents. Numerical solutions to growth equations are given. A solution to growth
coupled with nucleation is also described. Dissolution and surface reaction are dis-
cussed in Chapter 17. Formation of clouds by condensation is discussed further in
Chapter 18.
16.1 FLUXES TO AND FROM A SINGLE DROP
Condensation occurs when a gas diffuses to a homogeneously or heterogeneously
nucleated particle surface and changes state to a liquid on the surface. Common
condensable gases include water vapor, sulfuric acid, and by-products of toluene,
xylene, alkylbenzene, alkanes, alkenes, and biogenic hydrocarbon oxidation (e.g.,
Pandis et al. 1992). Solid deposition occurs when a gas diffuses to and changes
state to a solid on a nucleated surface. Deposition of water vapor onto particles
to form ice is an important mechanism of ice crystal formation in the atmosphere.
Evaporation and sublimation are the reverse of condensation and ice deposition,
respectively. In this section, growth equations for the condensation of water vapor
onto a single homogeneous liquid water drop are derived.
FromFicks rst lawof diffusion, the rate of change of mass of a single, spherical,
homogeneous liquid water drop can be described with
dm
dt
= 4 R
2
D
v
d
v
dR
(16.1)
where m is the mass of the drop (g), R is the radial distance from the center of the
drop (cm), D
v
is the molecular diffusion coefcient of water vapor in air (cm
2
s
1
),
v
is the density of water vapor (g cm
3
), and d
v
dRis the radial gradient of vapor
density. Equation (16.1) applies to large particles in the continuumregime (Kn 1).
525
Condensation, evaporation, deposition, and sublimation
Integrating the equation from the drop surface, where
v
=
v.r
and R= r, to
innity, where
v
=
v
and R= , gives
dm
dt
= 4r D
v
(
v
v.r
) (16.2)
If
v
>
v.r
, vapor diffuses to and condenses on the drop surface. If
v
-
v.r
,
condensate evaporates from and diffuses away from the surface.
When water vapor condenses, latent heat is released, increasing the temperature
at the drop surface, creating a gradient between the surface and the surrounding air.
The gradient is reduced by the conduction of energy away from the drop surface
by air molecules. The equation for the cooling rate at the drop surface due to
conduction is
dQ
r
dt
= 4 R
2
a
dT
dR
(16.3)
(J s
1
), where
a
is the thermal conductivity of moist air (J cm
1
s
1
K
1
), and
the temperature gradient is negative since temperature decreases with increasing
distance from the drop during condensation. Integrating (16.3) from the drop
surface, where T = T
r
, to innity where T = T, gives
dQ
r
dt
= 4r
a
(T
r
T) (16.4)
Equations (16.2) and (16.4) were rst written in a similar form by James Clerk
Maxwell in 1877 (Maxwell 1890). The two equations are related to temperature
at the drop surface by
mc
W
dT
r
dt
= L
e
dm
dt

dQ
r
dt
(16.5)
where c
W
is the specic heat of liquid water (J g
1
K
1
), and L
e
is the latent heat
of evaporation of water (J g
1
). Combining (16.4) with (16.5) under steady-state
conditions (dT
r
dt = 0) gives
L
e
dm
dt
= 4r
a
(T
r
T) (16.6)
The remainder of this derivation originates from Mason (1971) (see also Rogers
and Yau 1989). Combining the equation of state at saturation, p
v.s
=
v.s
R
v
T, with
the ClausiusClapeyron equation, dp
v.s
dT =
v.s
L
e
T from (2.57), gives
d
v.s
v.s
=
L
e
R
v
dT
T
2

dT
T
(16.7)
Integrating this equation from innity, where T = T and
v.s
=
v.s
(T), to the drop
surface, where T = T
s
and
v.s
=
v.s
(T
r
), yields
ln

v.s
(T
r
)
v.s
(T)
=
L
e
R
v
(T
r
T)
TT
r
ln
T
r
T
(16.8)
526
16.1 Fluxes to and from a single drop
Since T T
r
, (16.8) simplies to
v.s
(T
r
)
v.s
(T)
v.s
(T)
=
L
e
R
v
(T
r
T)
T
2

T
r
T
T
(16.9)
Substituting T
r
T from (16.6) into (16.9) gives
v.s
(T
r
)
v.s
(T)
v.s
(T)
=
L
e
4r
a
T
L
e
R
v
T
1
dm
dt
(16.10)
Dividing (16.2) by
v.s
(T), rearranging, adding the result to (16.10), and assuming
v.r

v.s
(T
r
) results in
v.s
(T)
v.s
(T)
=
L
e
4r
a
T
L
e
R
v
T
1
+
1
4r D
v
v.s
(T)
dm
dt
(16.11)
Substituting
v.s
=
v.s
(T),
v.s
= p
v.s
R
v
T, and
v
= p
v
R
v
T into (16.11) and solv-
ing for dmdt gives the mass-ux form of the growth equation for a single, homo-
geneous liquid water drop as
dm
dt
=
4r D
v
( p
v
p
v.s
)
D
v
L
e
p
v.s
a
T
L
e
R
v
T
1
+ R
v
T
(16.12)
This equation can be modied to treat any gas by replacing L
e
, D
v
, p
v
, p
v.s
, and
R
v
with L
e.q
, D
q
, p
q
, p
q.s
, and R
m
q
, respectively, and to treat size dependence
by adding the subscript i to m and r. The molecular diffusion coefcient (D
q
) and
thermal conductivity (
a
) can be modied by considering the collision geometry
between vapor molecules and a particle surface, the probability that molecules
stick to a surface upon collision, and eddies created by the particle. The satura-
tion vapor pressure ( p
q.s
) can be modied by considering surface curvature, the
presence of solute on a surface, and radiative heating/cooling. The corrections will
be discussed shortly, but for now, a subscript i and a prime are added to D
q
,
a
,
p
q.s
to denote size dependence and a modied value, respectively. With the above
substitutions in (16.12), the mass-ux form of the growth equation of substance q
to one homogeneous drop of size i becomes
dm
i
dt
=
4r
i
D
q.i
( p
q
p
q.s.i
)
D
q.i
L
e.q
p
q.s.i
a.i
T
L
e.q
m
q
R
T
1
+
R
T
m
q
(16.13)
Sometimes, the assumption L
e.q
m
q
R
T 1 L
e.q
m
q
R
T is made. If T =273 K,
L
e.q
m
q
R
T 19.8 for water vapor, and the assumption induces an error of 1/19.8.
Because (16.13) is a better approximation, it is used hereafter.
The change in mass of a spherical drop is related to its change in radius by
dm
i
dt
= 4r
2
i

p.i
dr
i
dt
(16.14)
527
where
p.i
is the drop density (g cm
3
). Combining (16.14) with (16.13) gives the
radius ux form of the growth equation for a single, homogeneous, spherical drop
as
r
i
dr
i
dt
=
D
q.i
( p
q
p
q.s.i
)
D
q.i
L
e.q
p.i
p
q.s.i
a.i
T
L
e.q
m
q
R
T
1
+
R
T
p.i
m
q
(16.15)
The time rate of change of mass of a spherical drop is related to its change of vol-
ume by dm
i
dt =
p.i
d
i
dt, and radius is related to volume by r
i
= (3
i
4)
13
.
Combining these expressions with (16.13) gives the volume ux formof the growth
equation for a single, homogeneous, spherical drop as
d
i
dt
=
(48
2
i
)
13
D
q.i
( p
q
p
q.s.i
)
D
q.i
L
e.q
p.i
p
q.s.i
a.i
T
L
e.q
m
q
R
T
1
+
R
T
p.i
m
q
(16.16)
16.2 CORRECTIONS TO GROWTH PARAMETERS
In this section, corrections to the molecular diffusion coefcient, thermal conduc-
tivity of air, and the saturation vapor pressure used in the growth equation are
described.
16.2.1 Corrections to the molecular diffusion coefcient
Molecular diffusion is the movement of molecules due to their kinetic energy and
redirection of the molecules due to their collision with other molecules. Molecules
in a body with a temperature greater than 0 K have an average kinetic energy and
thermal speed given by (11.151). When a molecule moves, it collides with other
molecules, which redirect it along an arbitrary path. The pace at which a molecule
spreads by diffusion is proportional to its thermal speed and the distance between
collisions. Thermal speed is proportional to the square root of temperature, and
the distance between collisions is inversely proportional to the density of air. An
expression for the molecular diffusion coefcient of a trace gas in air (cm
2
s
1
)
with these characteristics is
D
q
=
5
16Ad
2
q
Tm
a
2
m
q
+m
a
m
q
(16.17)
(e.g., Chapman and Cowling 1970; Davis 1983), where A is Avogadros num-
ber (6.022 136 7 10
23
molec. mol
1
), d
q
is the collision diameter (cm) of gas
molecule q,
a
3
), R
is the universal gas constant (8.314 51

10
7
g cm
2
s
2
mol
1
K
1
), T is absolute temperature (K), m
a
is the molecular
weight of air (28.966 g mol
1
), and m
q
is the molecular weight of gas q (g mol
1
).
528
16.2 Corrections to growth parameters
Table 16.1 Collision diameters and diffusion coefcients of several gases at
288 K and 1013 hPa
Collision diameter Molecular weight Diffusion coefcient
Gas () (g mol
1
) (cm
2
s
1
)
Air 3.67 28.966 0.147
Ar 3.58 39.948 0.144
CO
2
4.53 44.011 0.088
H
2
2.71 2.014 0.751
He 2.15 4.002 0.871
Xe 4.78 131.29 0.067
Kr 4.08 83.80 0.098
N
2
3.70 28.013 0.146
NH
3
4.32 17.031 0.123
Ne 2.54 20.180 0.339
O
2
3.54 31.999 0.154
H
2
O 3.11 18.015 0.234
Collision diameters are from Lide (2003), except that for water vapor, the colli-
sion diameter is derived by equating (16.17) with (8.14). Air density at the given
temperature and pressure is 0.001 23 g cm
3
.
Equation (16.17) simplies to the expression for the kinematic viscosity of air from
(4.55)(4.54) when the gas considered is air. Table 16.1 gives collision diameters
and diffusion coefcients for several gases from (16.17). Example 16.1 shows that
the diffusion coefcient increases with increasing altitude (lower pressure).
Example 16.1
Calculate the molecular diffusion coefcient of carbon dioxide in air when the
temperature is T = 220 K and air pressure is p
a
= 25 hPa. Assume the air is
dry.
SOLUTION
Since the air is dry, R
m
= R
. Fromthe equation of state, (2.36),

a
=0.000 039 6
g cm
3
. From Table 16.1, the molecular weight and molecular diameter of
carbon dioxide are m
q
= 44.011 g mol
1
and 4.53 10
8
cm, respectively.
Substituting these values into (16.17) gives the molecular diffusion coefcient
of CO
2
as D
q
= 2.39 cm
2
s
1
. Comparing this result with that in Table 16.1
suggests that the diffusion coefcient increases with decreasing air pressure.
Near a particle surface, the diffusion coefcient of a gas is affected by collision
geometry, sticking probability, and ventilation. A corrected molecular diffusion
coefcient (cm
2
s
1
) that takes these effects into account is
D
q.i
= D
q
q.i
F
q.L.i
(16.18)
529
where
q.i
is the correction for collision geometry and sticking probability, and
F
q.i
is the correction for ventilation.
The correction for collision geometry accounts for noncontinuum-regime
growth and is required, since (16.1) applies only to the continuum regime. The
correction is a function of the ratio of the distance between molecular collisions
of a diffusing gas and the size of the particle it is diffusing to. The correction for
sticking probability takes account of the ability of a gas to stick to a particle sur-
face once the gas has diffused to the surface. An expression for the two corrections,
together, is
q.i
=
1 +
1.33 +0.71Kn
1
q.i
1 +Kn
1
q.i
+
4(1
q.i
)
3
q.i
Kn
q.i
1
(16.19)
(Fuchs and Sutugin 1971; Pruppacher and Klett 1997), where
Kn
q.i
=

q
r
i
(16.20)
is the Knudsen number of the condensing gas with respect to particles of size i.
This Knudsen number gives the ratio of the mean free path of the condensing gas
to the size of the particle of interest. The mean free path of a gas molecule is the
average distance it travels before colliding and exchanging momentum with any
other gas molecule.
If the Knudsen number is large, the particle of interest is small relative to the
distance between molecular collisions, and the particle is intercepted relatively
infrequently by gas molecules. If the Knudsen number is small, the particle is large
relative to the distance between molecular collisions, and the particle is frequently
intercepted by gas molecules. In sum, (16.19) states that, in the absence of sticking
probability,
q.i

0 as Kn
q.i
(small particles)
1 as Kn
q.i
0 (large particles)
(16.21)
In a dilute mixture, where the number concentration of a trace-gas is much less
than that of air, the mean free path of a gas molecule is
q
=
m
a
Ad
2
q
m
a
m
a
+m
q
(16.22)
(Jeans 1954; Davis 1983). In (16.22), the mean free path is inversely proportional
to the air density. The greater the air density, the shorter the distance between
collisions of a gas molecule with an air molecule. Combining (16.17) and (16.22)
with the thermal speed of a gas molecule ( v
q
, cm s
1
) from (11.151) gives
q
=
64D
q
5
q
m
a
m
a
+m
q
(16.23)
In (16.19),
q,i
is the mass accommodation (sticking) coefcient of the gas, which
is the fractional number of collisions of gas q with particles of size i that results in the
530
gas sticking to the surface of the particle. Mass accommodation coefcients differ
from reaction probabilities, discussed in Chapter 11, in that reaction probabilities
take account of molecular diffusion to and reaction with a material on a particle
surface. Mass accommodation coefcients take account of only adsorption of a
gas to the surface.
Accommodation coefcients for soluble species range from 0.01 to 1.0
(Mozurkewich et al. 1987; Jayne et al. 1990; Van Doren et al. 1990; Chameides
and Stelson 1992). Van Doren et al. (1990) found that the mass accommodation
coefcient for nitric acid (HNO
3
) on liquid water varied from 0.07 at 268 K to
0.193 at 293 K, the coefcient for hydrochloric acid (HCl) on liquid water varied
from0.064 at 294 Kto 0.177 at 274 K, and the coefcient for dinitrogen pentoxide
(N
2
O
5
) on liquid water varied from 0.04 at 282 K to 0.061 at 271 K.
To correct for the increased rate of vapor and energy transfer to the upstream
surface of a large, falling particle, a ventilation factor for vapor is included in
the corrected molecular diffusion coefcient. The ventilation factor is the vapor
mass ux to (or from) a particle when it is moving, divided by the ux when it is at
rest. Increased vapor transfer occurs when a large particle falls, creating eddies that
sweep additional vapor toward the particle. The ventilation factor is small for small,
falling particles because such particles move too slowly to create eddies. An expres-
sion for the ventilation factor for a gas q interacting with a liquid water drop is
F
q.L.i
=
1 +0.108x
2
q.i
x
q.i
1.4
0.78 +0.308x
q.i
x
q.i
> 1.4
x
q.i
= Re
i
12
Sc
13
q
(16.24)
(Pruppacher and Klett 1997). The ventilation factor depends on the particle
Reynolds number, dened in (15.26), and the gas Schmidt number,
Sc
q
=

a
D
q
(16.25)
which is the ratio of the kinematic viscosity of air to the molecular diffusion coef-
cient of a gas. Since the particle Reynolds number is proportional to a particles
radius and its fall speed, the larger a particle and the faster it falls, the greater the
ventilation factor.
16.2.2 Corrections to the thermal conductivity of air
Conductive energy transfer to and from a particle surface is affected by some of
the same correction factors that affect vapor transfer. These factors are embodied
in the corrected thermal conductivity term
a.i
=
a
h.i
F
h.L.i
(16.26)
where
a
is the uncorrected thermal conductivity of moist air (J cm
1
s
1
K
1
),
531
from (2.7),
h.i
is the correction factor for collision geometry and sticking proba-
bility, and F
h.L.i
is the thermal ventilation factor. The correction factor for collision
geometry and sticking probability is
h.i
=
1 +
1.33 +0.71Kn
1
h.i
1 +Kn
1
h.i
+
4(1
h
)
3
h
Kn
h.i
1
(16.27)
where
Kn
h.i
=

h
r
i
(16.28)
is the Knudsen number for energy with respect to particles of size i. In this equation,
h
is the thermal mean free path (cm), which is the average distance that an air
molecule travels before exchanging energy by conduction with another molecule.
It is quantied as
h
=
3D
h
v
a
(16.29)
where D
h
is the molecular thermal diffusivity, dened in (8.14), and v
a
is the thermal
speed of an air molecule, dened in (15.25). The molecular thermal diffusivity
depends on the thermal conductivity of air, the specic heat of air, and the air
density.
In (16.27),
h
is the thermal accommodation coefcient, interpreted as the frac-
tional number of molecules bouncing off the surface of a drop that have acquired
the temperature of the drop (Pruppacher and Klett 1997). An equation for
h
is
h
=
T
m
T
T
s
T
(16.30)
where T
m
is the temperature of vapor molecules leaving the surface of a drop (K),
T
s
is the temperature of the surface (K), and T is the temperature of the ambient
vapor (K). For water-vapor growth,
h
has a typical value of 0.96.
The thermal ventilation factor has a meaning similar to that for vapor. When a
large particle falls through the air, it creates eddies that sweep additional energy to
the surface of the particle. An empirical equation for the thermal ventilation factor,
obtained from experiments with liquid water drops, is
F
h.L.i
=
1 +0.108x
2
h.i
x
h.i
1.4
0.78 +0.308x
h.i
x
h.i
> 1.4
x
h.i
= Re
12
i
Pr
13
(16.31)
(Pruppacher and Klett 1997), where the dimensionless Prandtl number,
Pr =

a
c
p.m
a
(16.32)
532
is proportional to the ratio of the dynamic viscosity of air to its thermal conduc-
tivity.
16.2.3 Corrections to the saturation vapor pressure
Expressions for the saturation vapor pressure of water over dilute, at liquid water
and ice surfaces were given in (2.62) and (2.65), respectively. Curvature, solutes,
and radiative cooling modify these saturation vapor pressures.
16.2.3.1 Curvature effect
The saturation vapor pressure increases over a curved surface relative to over a
at surface due to the curvature (Kelvin) effect. The surface of a small spherical
particle is more curved than that of a large spherical particle, causing the saturation
vapor pressure over a small particle to be greater than that over a large particle.
The saturation vapor pressure over a curved surface is greater than that over a at
surface because molecules desorb more readily from a curved surface than from a
at surface. Surface tension also plays a role in the curvature effect. The greater the
surface tension of a particle against air, the more likely a molecule desorbs from
the surface, and the greater the saturation vapor pressure.
The saturation vapor pressure over a curved, dilute surface relative to that over
a at, dilute surface is
p
q.s.i
p
q.s
= exp
2
p
m
p
r
i
R
T
p.i
1 +
2
p
m
p
r
i
R
T
p.i
(16.33)
where p
q.s.i
is the saturation vapor pressure of gas q over a curved, dilute surface
(hPa), p
q.s
is the saturation vapor pressure over a at, dilute surface (hPa),
p
is
the average particle surface tension (dyn cm
1
= g s
2
), m
p
is the average particle
molecular weight (g mol
1
), r
i
is particle radius (cm), R
is the universal gas con-

stant (8.31451 10
7
g cm
2
s
2
mol
1
K
1
), T is absolute temperature (K), and
p,i
is the average particle density (g cm
3
). The second expression was obtained
by noting that the exponent in the rst term is small, and e
x
1 + x for small x.
For liquid water and ice, p
q.s
= p
v.s
, and p
q.s
= p
v.I
, respectively, equations for
which were given in Section 2.5.3. For sulfuric acid, parameterizations of the sat-
uration vapor pressures are available from Bolsaitis and Elliott (1990) and Vehka-
maki et al. (2002). Vapor pressure information for several organic gases can be
obtained from McMurry and Grosjean (1985), Tao and McMurry (1989), Odum
et al. (1996), Makar (2001), and Makar et al. (2003).
An expression for the surface tension of pure liquid water against air (
p
=
wa
)
was given in (14.19). The surface tension of water containing dissolved organic
material is less than that of pure water (e.g., Li et al. 1998; Facchini et al. 1999).
An empirical expression for the surface tension of water containing dissolved
533
organics is
p
=
wa
0.0187T ln(1 +628.14m
C
) (16.34)
(Facchini et al. 1999), where
p
is in units of dyn cm
1
(g s
2
), T is absolute
temperature (K), and m
C
is the molality of carbon dissolved in water (moles carbon
per kilogram of water). Aerosol particles often also contain dissolved inorganic
solutes, such as sodium chloride, ammonium sulfate, potassium nitrate, etc. These
solutes frequently dissociate into ions, such as Na
+
, Cl
; NH
4
+
, SO
4
2
; K
+
, NO
3
,
respectively. Dissolved inorganic ions have the opposite effect to dissolved organics;
they increase the surface tension relative to pure water. An empirical expression
for the surface tension of water containing dissolved inorganic ions is
p
=
wa
+1.7m
I
(16.35)
(e.g., Pruppacher and Klett 1997), where
p
is in units of dyn cm
1
(g s
2
) and m
I
is the molality of total dissolved ions in water. Thus, if a solute, such as sodium
chloride, dissociates, m
I
is the molality of the sodium ion plus that of the chloride
ion (or twice the molality of undissociated sodium chloride). Equations (16.34)
and (16.35) suggest that the addition of dissolved organics decreases the saturation
vapor pressure and the addition of dissolved ions increases the saturation vapor
pressure of water over a curved surface when only the effect of surface tension on
the curvature effect is considered.
Example 16.2
Compare the surface tensions of a (a) pure water drop, (b) water drop con-
taining 0.03 mol-C kg
1
of an undissociated organic compound, and (c) water
drop containing 0.03 mol kg
1
of dissociated ammonium sulfate (NH
4
)
2
SO
4
at
T = 298.15 K.
SOLUTION
From (14.19), (16.34), and (16.35), the surface tensions in the three respective
cases are
wa
= 76.1 0.155 25
C = 72.225 dyncm
1
C
=
wa
0.0187 298.15 K ln(1 +628.14 0.03 mol-Ckg
1
)
= 55.56 dyncm
1
I
=
wa
+1.7 3 0.03 mol kg
1
= 72.378 dyncm
1
The factor of three in the last equation accounts for the fact that ammonium
sulfate dissociates into three ions. The organic compound reduced the surface
tension of water, whereas the dissolved ions increased it.
534
16.2.3.2 Solute effect
A second factor that affects the saturation vapor pressure is that of dissolved solute
(dissociated or undissociated). When a solute dissolves in solution, the saturation
vapor pressure of the solvent is reduced to the saturation vapor pressure of the
pure solvent multiplied by its mole fraction in solution. Thus, if n
s
is the num-
ber of moles of solute dissolving in n
w
moles of water (the solvent) in a liquid
drop of size i , the saturation vapor pressure of water (species q) over the solution
surface is
p
q.s.i
p
q.s
=
n
w
n
w
+n
s
(16.36)
This is known as Raoults law (Raoult 1887) (also called the solute effect), named
after Francois-Marie Raoult who studied the vapor pressures and freezing points of
solutions. The reason for the reduction in the solvents vapor pressure is that solute
molecules replace solvent molecules on the solution surface, preventing solvent
molecules from evaporating. According to Raoults law, the vapor pressure of the
dissolved solute over the solution similarly equals the vapor pressure of the pure
solute multiplied by its mole fraction in solution. Thus, although the solvents
vapor pressure is reduced, it is also replaced by solute vapor pressure. Thus, the
total vapor pressure over the solution is the sum of the vapor pressures of the
solvent and solute over the solution. As such, if the vapor pressure of a pure solute
is zero (e.g., if the solute is involatile it does not evaporate), the total vapor
pressure over the solution is less than that of the pure solvent alone.
For dilute solutions, n
w
n
s
, and (16.36) simplies to
p
q.s.i
p
q.s
1
n
s
n
w
(16.37)
The number of moles of solute in solution is approximately n
s
= i
v
M
s
m
s
, where
M
s
is the mass of solute in the particle (g), m
s
is the molecular weight of the solute
(g mol
1
), and i
v
is the degree of dissociation of the solute into ions, called the
vant Hoff factor. The factor i
v
gives the actual number of moles of ions that
dissociate from one mole of solute. For sodium chloride (NaCl), which dissociates
in solution to Na
+
and Cl
, i
v
= 2. For ammonium sulfate ((NH
4
)
2
SO
4
), which
dissociates to 2NH
4
+
and SO
4
2
, i
v
= 3. For a nondissociating solute, i
v
= 1. The
more ions present on the surface, the greater the reduction in saturation vapor
pressure.
The number of moles of liquid water in a drop, used in (16.37), is approximately
n
w
=
M
w
m
v
4r
3
i

w
3m
v
(16.38)
where M
w
is the mass of liquid water in the drop (g), m
v
is the molecular weight
of water (g mol
1
), and
w
is the density of liquid water (g cm
3
). This equation
535
assumes that M
w
M
s
. Substituting n
s
= i
v
M
s
m
s
and (16.38) into (16.37) gives
the solute effect over a dilute liquid water solution as
p
q.s.i
p
q.s
1
3m
v
i
v
M
s
4r
3
i

w
m
s
(16.39)
The solute effect affects primarily small particles, since small particles have a higher
concentration of solute than do large particles. Equation (16.39) can be modied
to treat multiple dissolved gases simultaneously, as described in Section 18.8.1.
16.2.3.3 K ohler equation
Taking the product of the right sides of (16.33) and (16.39) and eliminating the
last term of the result, which is small, gives the K ohler equation (K ohler 1936),
S
q.i
=
p
q.s.i
p
q.s
1 +
2
p
m
p
r
i
R
T
p.i
3m
v
i
v
M
s
4r
3
i

w
m
s
(16.40)
where S
q.i
is the saturation ratio at equilibrium. The K ohler equation relates the
saturation vapor pressure of water over a curved surface containing solute to that
over a at surface without solute.
The K ohler equation is useful for determining whether a cloud drop can form
from an aerosol particle under ambient conditions. All cloud drops in the atmo-
sphere form on aerosol particles. As discussed in Section 14.2.1, water vapor can-
not homogeneously nucleate under atmospheric conditions but can readily het-
erogeneously nucleate on an existing surface. The most prevalent surfaces in the
atmosphere are aerosol particle surfaces. Following heterogeneous nucleation on
an aerosol particle, water vapor readily grows onto the particle if the air is super-
saturated with water over the particles surface.
Water does not easily heterogeneously nucleate on all aerosol particle surfaces.
Those particles that water can heterogeneously nucleate on to formliquid drops are
called cloud condensation nuclei (CCN). Most aerosol particles can serve as a CCN
although some, such as diesel soot particles (mostly black carbon plus lubricating
oil), are relatively hydrophobic, so water does not readily condense upon them
(e.g., Ishizaka and Adhikari 2003). As soot ages, though, other chemicals coat it
through condensation and coagulation. If the coating chemicals are soluble (e.g.,
sulfuric acid, nitric acid, hydrochloric acid, hydroscopic organic matter), soots
ability to act as a CCN increases.
The heterogeneous nucleation and growth of a CCN to a cloud drop is called
CCN activation. Whether a single CCN of a given size activates into a cloud drop
depends substantially on three parameters: the ambient saturation ratio, the critical
saturation ratio, and the critical radius for growth. The rst parameter depends on
ambient conditions, and the last two are derived from the K ohler equation.
536
0.98
0.99
1
1.01
1.02
0.01 0.1 1 10
S
a
t
u
r
a
t
i
o
n

r
a
t
i
o
Particle radius (m)
Curvature effect
Solute effect
Equilibrium
saturation
ratio
r*
S*
Figure 16.1 Example of how curvature and
solute effects affect the equilibrium saturation
ratio S
q.i
. The sum of the curvature and solute
curves (dashed line) is S
q.i
. The curvature effect
increases the saturation vapor pressure over
small aerosol particles, increasing S
q.i
. The
solute effect decreases the saturation vapor
pressure over small aerosol particles, decreas-
ing S
q.i
. The maximum S
q.i
is the critical sat-
uration ratio S
, which occurs at the critical

radius r
.
The ambient saturation ratio of a gas is the ratio of the partial pressure of the
gas ( p
q
) to the gass saturation vapor pressure over a at, dilute surface ( p
q.s
):
S
q
=
p
q
p
q.s
(16.41)
When S
q
> S
q.i
, the air is supersaturated with respect to particles of size i, and vapor
condenses onto these particles. Figure 16.1, though, shows that the equilibrium
saturation ratio S
q.i
(from 16.40) varies with particle radius. Thus, although S
q
>
S
q.i
initially in many cases, the condition does not necessarily hold as the particle
grows. Suppose the ambient saturation ratio S
q
=1.001 and a particle is small (e.g.,
- 0.1 m in radius). At that radius, S
q
> S
q.i
, and the particle starts to grow. As
it grows, S
q.i
increases along the equilibrium saturation ratio curve in Fig. 16.1.
When the particle exceeds about 0.15 m radius, S
q
S
q.i
and the particle cannot
grow further. If, instead, S
q
were about 1.003, the particle would grow until it
reached cloud-drop size (about 510 m). Alternatively, if the initial particle were
large (e.g., > 0.6 m in radius) and S
q
= 1.001, the particle would also grow to
cloud-drop size.
In Fig. 16.1, the radius at which the equilibrium saturation ratio is maximum is
the critical radius for growth (r
). The equilibrium saturation ratio at this radius is

the critical saturation ratio (S
). Aerosol particles that start smaller than the critical

radius cannot grow (activate) to cloud-drop size unless the ambient saturation
ratio exceeds the critical saturation ratio. Aerosol particles initially larger than the
critical radius can activate if S
q
> S
q.i
over such drops. In sum, activation of a
CCN can occur if r
i
- r
and S
q
> S
or r
i
> r
and S
q
> S
q.i
. When r
i
- r
and
S
> S
q
> S
q.i
, a CCN cannot activate, and it grows no larger than r
. Also, when
537
Table 16.2 Critical radii for growth and critical supersaturations for aerosol
particles containing dissolved sodium chloride or ammonium sulfate when
solute mass varies at 275 K. The effect of solute on surface tension is ignored.
Sodium chloride Ammonium sulfate
Solute mass (g) r
(m) S
1 (%) r
(m) S
1 (%)
0 0 0
10
18
0.019 4.1 0.016 5.1
10
16
0.19 0.41 0.16 0.51
10
14
1.9 0.041 1.6 0.051
10
12
19 0.0041 16 0.0051
S
q
- S
q.i
, a CCN does not activate. When r
i
> r
, large particles are more likely

to grow than small particles because the growth rate is proportional to S
q
S
q.i
,
and this difference is larger for large particles than for small particles.
The critical radius and critical saturation ratio are found by rewriting (16.40)
as
S
q.i
= 1 +
a
r
i
b
r
3
i
a =
2
p
m
p
R
T
p.i
b =
3m
v
i
v
M
s
4
w
m
s
(16.42)
where a has units of cmand b has units of cm
3
. Taking the derivative of S
q.i
, setting
the derivative to zero, solving for the radius, and substituting the result back into
(16.42) gives the critical radius for growth and critical saturation ratio as
r
3b
a
S
= 1 +
4a
3
27b
(16.43)
respectively.
Supersaturation is the saturation ratio minus one. Table 16.2 gives critical radii
and critical supersaturations (expressed as percentages) when liquid particles con-
tain either dissolved sodium chloride or dissolved ammonium sulfate. In the atmo-
sphere, ambient supersaturations rarely exceed 2 percent (saturation ratio of 1.02)
and are usually much smaller. Since the number concentration of cloud drops (100
1000 cm
3
) is less than that of CCN (10
3
10
6
cm
3
), only a fraction of CCN
activate into cloud drops. In the absence of solute, r
=0, so all particles for which

S
q
> S
q.i
have potential to activate to cloud drops.
16.2.3.4 Radiative cooling effect
When a large water-containing particle releases thermal-infrared radiation, it cools,
reducing the temperature-dependent saturation vapor pressure of water over its
surface. Large water-containing particles also warm by absorbing solar radiation,
but their thermal-infrared cooling rates are generally faster than are their solar
538
heating rates. Small liquid water particles do not absorb much solar radiation,
because their solar absorption efciencies are negligible, as shown in Fig. 9.21.
The net change in saturation vapor pressure over a liquid drop surface due to
radiative cooling or heating is called the radiative cooling effect since cooling is the
predominant result of this effect.
The saturation vapor pressure over a at, dilute surface that experiences net
radiative cooling relative to that over a at, dilute surface that does not is
p
q.s.i
p
q.s
1 +
L
e.q
m
q
H
r.i
4r
i
R
T
2
d.i
(16.44)
where
H
r.i
=
r
2
i

0
Q
a
(m
.
i.
)(I
)d (16.45)
is the radiative cooling rate (W) of the particle (e.g., Toon et al. 1989b). In this
equation, Q
a
is the single-particle absorption efciency, I
is the incoming spectral

radiance, and B
is the Planck spectral radiance emitted by the particle surface.

Absorption efciencies, which depend on the complex index of refraction and size
parameter (m
and
i.
), were discussed in Section 9.4.3. A method of determin-
ing radiance was given in Section 9.8.3. The Planck function was given in (9.4).
Radiative cooling rates can be positive or negative.
16.2.3.5 Overall effects
For condensation of water vapor onto a liquid-water surface, the overall equilib-
rium saturation ratio is obtained by taking the product of the right sides of (16.40)
and (16.44) and eliminating small terms. The result is
S
q.i
=
p
q.s.i
p
q.s
1 +
2
p
m
p
r
i
R
T
p.i
3m
v
i
v
M
s
4r
3
i

w
m
s
+
L
e.q
m
q
H
r.i
4r
i
R
T
2
d.i
(16.46)
The solute-effect term in (16.46) is most applicable to water vapor growth onto
liquid drops. For growth of vapor onto small particles, the radiative cooling term
is ignored, since small particles have masses too small to affect the saturation
vapor pressure noticeably through this term. Since most gases other than water
vapor grow primarily onto small particles, the term is also ignored for such gases.
Removing the solute and radiative terms from (16.46) gives an estimate of the
overall equilibrium saturation ratio for the growth of gases other than water vapor
as
S
q.i
=
p
q.s.i
p
q.s
1 +
2
p
m
p
r
i
R
T
p.i
(16.47)
539
16.3 FLUXES TO A PARTICLE WITH MULTIPLE
COMPONENTS
The equations in Section 16.1 give the change in mass, volume, and radius of
a single, homogeneous particle. With one component present, the volume of a
spherical drop changes proportionately to the one-third power of the volume of
the condensed material already in the drop, as shown in (16.16). When multiple
components are present within a drop, the volume changes proportionately to the
one-third power of the summed volume of all components. The growth equation
for individual components in a drop containing multiple components is derived
below.
The total volume and mass of one particle in which one component grows can
be dened as
i.t
=
q.i.t
+
i.th
q.i.th
(16.48)
m
i.t
=
p.i.t
i.t
=
p.q
q.i.t
+
p.i.th
i.th
p.q
q.i.th
(16.49)
respectively, where
q.i.t
is the volume of growing component q in size bin i at
current time t,
i.th
is the initial total volume of the particle before growth,
q.i.th
is the initial volume of component q in the particle before growth (time t h),
p.i.th
is the mass density (g cm
3
) of the total particle before growth, and
p.q
3
) of component q, which is assumed to be constant. In
reality, the mass density of a component in a particle may vary. The time derivative
of (16.49) is
dm
i.t
dt
=
p.i.t
d
i.t
dt
=
p.q
d
q.i.t
dt
(16.50)
where d
i.th
dt = d
q.i.th
dt = 0. Combining (16.50) and (16.48) with (16.16)
gives the volume rate of change of one component in a multicomponent particle
as
d
q.i.t
dt
=
[48
2
(
q.i.t
+
i.th
q.i.th
)]
13
D
q.i
( p
q
p
q.s.i
)
D
q.i
L
e.q
p.q
p
q.s.i
a.i
T
L
e.q
m
q
R
T
1
+
R
T
p.q
m
q
(16.51)
16.4 FLUXES TO A POPULATION OF PARTICLES
Equation (16.51) gives the ux of one gas to one particle that contains many com-
ponents. The uxes of individual gases to many particles, each with different size
and composition, are now considered. In a model, particles are often segregated
into size bins, and each particle in a bin is assumed to have the same volume and
composition as each other particle in the bin. The volume (cm
3
) of a single particle
is related to its volume concentration (cm
3
cm
3
) and number concentration (par-
ticles cm
3
) by
i
= v
i
n
i
. The volume of an individual component in a particle
is related to its volume concentration and the number concentration of particles
by
q.i.t
= v
q.i.t
n
i.th
. Substituting expressions of this type into (16.51) gives the
540
16.4 Fluxes to a population of particles
volume-concentration form of the growth equation for a component in particles
of size i as
dv
q.i.t
dt
=
n
23
i.th
[48
2
(v
q.i.t
+v
i.th
v
q.i.th
)]
13
D
q.i
( p
q
p
q.s.i
)
D
q.i
L
e.q
p.q
p
q.s.i
a.i
T
L
e.q
m
q
R
T
1
+
R
T
p.q
m
q
(16.52)
(Jacobson and Turco 1995). Equation (16.52) can be simplied by dening
p
q
= C
q
R
T p
q.s.i
= C
q.s.i
R
T (16.53)
where C
q
is the gas-phase mole concentration (mol cm
3
) of component q, and C
q.s.i
is an effective saturation vapor concentration (mol cm
3
). Combining (16.52) with
(16.53) and adding more time subscripts give
dv
q.i.t
dt
= n
23
i.th
[48
2
(v
q.i.t
+v
i.th
v
q.i.th
)]
13
D
eff
q.i.th
m
q
p.q
(C
q.t
C
q.s.i.th
)
(16.54)
where
D
eff
q.i.th
=
D
q.i
m
q
D
q.i
L
e.q
C
q.s.i.th
a.i
T
L
e.q
m
q
R
T
1
+1
(16.55)
is an effective molecular diffusion coefcient (cm
2
s
1
). For water vapor, C
q.s.i.th
is
large, and the left side of the denominator of (16.55) cannot be ignored. For other
gases, C
q.s.i.th
is usually orders of magnitude smaller than for water vapor, and the
left side of the denominator is much less than 1. Thus, for gases other than water
vapor, the left side of the denominator can be ignored, and (16.55) simplies to
D
eff
q.i.th
D
q.i
= D
q
q.i
F
q.L.i
=
D
q
F
q.L.i
1 +
1.33 +0.71Kn
1
q.i
1 +Kn
1
q.i
+
4(1
q.i
)
3
q.i
Kn
q.i
(16.56)
Equation (16.54) is the growth equation for a single gas to a population of spher-
ical particles of size i. The equation yields the change in volume concentration of
a particle with many components when one component in the particle grows or
evaporates. The equation is applied to each particle size bin and each volatile com-
ponent q. Since gases condense on particles in several size bins, a gas conservation
equation must be written to accompany (16.54). Such an equation is
dC
q.t
dt
=
p.q
m
q
N
B
i =1
dv
q.i.t
dt
(16.57)
541
16.5 SOLUTIONS TO GROWTH EQUATIONS
Many methods have been developed to solve time-dependent condensation and
evaporation equations. These include nite-element methods (Varoglu and Finn
1980; Tsang and Brock 1986; Tsang and Huang 1990), discrete-size-bin meth-
ods (e.g., Gelbard and Seinfeld 1980; Toon et al. 1988; Rao and McMurry 1989;
Lister et al. 1995; Gelbard et al. 1998), the cubic spline method (e.g., Middleton
and Brock 1976; Nguyen and Dabdub 2001), modied upwind difference methods
(e.g., Smolarkiewicz 1983; Tsang and Korgaonkar 1987), trajectory-grid methods
(e.g., Chock and Winkler 2000; Gaydos et al. 2003), semi-Lagrangian methods
(e.g., Bott 1989; Nguyen and Dabdub 2002), and moment methods (e.g., Fried-
lander 1983; Lee 1985; Whitby 1985; Brock et al. 1986; Binkowski and Shankar
1995). Some of these methods conserve mass between the gas and particle phases,
and others do not. Some methods reduce or eliminate numerical diffusion during
growth, and others are more diffusive. Some methods are computationally more
efcient than others. Zhang et al. (1999) compared several condensation schemes.
Below, two solutions to condensation/evaporation equations are given. Both
solutions conserve mass between the gas and aerosol phases and reduce or eliminate
numerical diffusion during growth when used in conjunction with a hybrid, full-
moving, or moving-center size structure. Although the rst method gives exact
solutions, it is iterative and requires more computational time than the second.
Both schemes have been used in three-dimensional models. The second method,
which is noniterative but unconditionally stable, is useful when computer time is
limited.
16.5.1 Integrated numerical solution
Together, (16.54) and (16.57) constitute N
B
+ 1 nonlinear, rst-order ordinary
differential equations (ODEs). A method of solving these equations is to integrate
them with a sparse-matrix ODE solver, such as that discussed in Section 12.8. An
advantage of such a solver is that the matrix of partial derivatives arising from
(16.54) and (16.57) is sparse. When N
B
= 3, a matrix of partial derivatives arising
from this set of equations is
v
q.1.t
v
q.2.t
v
q.3.t
C
q.t
v
q.1.t
v
q.2.t
v
q.3.t
C
q.t
1 h
s
2
v
q.1.t
v
q.1.t
t
0 0 h
s
2
v
q.1.t
C
q.t
t
0 1 h
s
2
v
q.2.t
v
q.2.t
t
0 h
s
2
v
q.2.t
C
q.t
t
0 0 1 h
s
2
v
q.3.t
v
q.3.t
t
h
s
2
v
q.3.t
C
q.t
t
h
s
2
C
q.t
v
q.1.t
t
h
s
2
C
q.t
v
q.2.t
t
h
s
2
C
q.t
v
q.3.t
t
1 h
s
2
C
q.t
C
q.t
t
(16.58)
542
16.5 Solutions to growth equations
Table 16.3 Reduction in array space and number of matrix
operations before and after sparse-matrix techniques are
applied to growth/evaporation ODEs
Value
After sparse
Quantity Initial reductions
Order of matrix (N
B
+ 1) 17 17
No. of initial matrix positions lled 289 49
Percent of initial positions lled 100 17
No. of nal matrix positions lled 289 49
Percent of nal positions lled 100 17
No. of operations in decomp. 1 1496 16
No. of operations in decomp. 2 136 16
No. of operations in backsub. 1 136 16
No. of operations in backsub. 2 136 16
The last four rows show the number of operations in each of four
loops of decomposition and backsubstitution.
The partial derivatives of this matrix are
2
v
q.i.t
v
q.i.t
t
=
1
3
n
i.th
v
q.i.t
23
48
2
13
D
eff
q.i.th
m
q
p.q
(C
q.t
C
q.s.i.th
) i = 1. . . . . N
B
(16.59)
2
v
q.i.t
C
q.t
t
= n
23
i.th
[48
2
(v
q.i.t
+v
i.th
v
q.i.th
)]
13
D
eff
q.i.th
m
q
p.q
i = 1. . . . . N
B
(16.60)
2
C
q.t
v
q.i.t
t
=
p.q
m
q
2
v
q.i.t
v
q.i.t
t
i = 1. . . . . N
B
(16.61)
2
C
q.t
C
q.t
t
=
p.q
m
q
N
B
i =1
2
v
q.i.t
C
q.t
t
(16.62)
Matrix (16.58) does not require reordering, since it is already ordered from
the fewest partial-derivative entries at the top of the matrix to the most at the
bottom. Table 16.3 conrms that the current ordering of the matrix is optimal.
It shows statistics from a case with 16 size bins (N
B
+1 = 17). The table shows
that no matrix ll-in was required during matrix decomposition and backsubsti-
tution.
An ODE integrator gives an accurate and relatively rapid solution to the
growth equations if sparse-matrix techniques are used. An integrator still requires
iteration; thus, in a model with 10
4
10
6
grid cells, faster solutions may be
desired.
543
16.5.2 Analytical predictor of condensation
Another method of solving growth equations is the analytical predictor of conden-
sation (APC) scheme (Jacobson 1997c, 2002). This scheme requires no iteration,
conserves mass exactly, and is unconditionally stable. The solution is obtained
by holding the volume on the right side of (16.54) constant for a time step.
Thus,
dv
q.i.t
dt
= n
23
i.th
(48
2
v
i.th
)
13
D
eff
q.i.th
m
q
p.q
(C
q.t
C
q.s.i.th
) (16.63)
This equation is modied by dening a mass-transfer coefcient (s
1
), an effective
saturation vapor mole concentration, and particle volume concentration as
k
q.i.th
= n
23
i.th
(48
2
v
i.th
)
13
D
eff
q.i.th
= n
i.th
4r
i.th
D
eff
q.i.th
(16.64)
C
q.s.i.th
= S
q.i.th
C
q.s.i.th
(16.65)
v
q.i.t
= m
q
c
q.i.t
p.q
(16.66)
respectively. In (16.65), C
q.s.i.th
= p
q.s.th
R
T is an uncorrected saturation vapor

mole concentration (mol cm
3
), and S
q.i.th
is the equilibrium saturation ratio of
the condensing gas. In (16.66), c
q.i.t
is the particle-phase mole concentration of
component q (moles per cubic centimeter of air). Equilibrium saturation ratios
were dened for liquid water in (16.46) and other gases in (16.47). Substituting
(16.64)(16.66) into (16.63) and (16.57) gives
dc
q.i.t
dt
= k
q.i.th
(C
q.t
S
q.i.th
C
q.s.i.th
) i = 1. . . . . N
B
(16.67)
dC
q.t
dt
=
N
B
i =1
[k
q.i.th
(C
q.t
S
q.i.th
C
q.s.i.t
)] (16.68)
respectively. Equations (16.67) and (16.68) represent N
B
+1 ODEs.
The noniterative solution to the growth equations is obtained by integrating
(16.67) for a nal aerosol concentration. The result is
c
q.i.t
= c
q.i.th
+hk
q.i.th
(C
q.t
S
q.i.th
C
q.s.i.th
) (16.69)
where the nal gas mole concentration, C
q.t
, is currently unknown. Final aerosol
and gas concentrations are constrained by the mass-balance equation,
C
q.t
+
N
B
i =1
(c
q.i.t
) = C
q.th
+
N
B
i =1
(c
q.i.th
) = C
tot
(16.70)
Substituting (16.69) into (16.70) and solving for the gas concentration give
C
q.t
=
C
q.th
+h
N
B
i =1
(k
q.i.th
S
q.i.th
C
q.s.i.th
)
1 +h
N
B
i =1
k
q.i.th
(16.71)
544
16.6 Solving homogeneous nucleation with condensation
Table 16.4 Demonstration of unconditional stability of the APC scheme
Time step size C
q,t
c
q,1,t
c
q,2,t
c
q,3,t
(s) (g m
3
) (g m
3
) (g m
3
) (g m
3
)
0.1 1.0 2.18 5.36 1.46
10 1.0 2.18 5.36 1.46
60 1.0 2.18 5.36 1.46
600 1.0 2.18 5.36 1.46
7200 1.0 2.18 5.36 1.46
Assume a gas transfers to and from three particle size bins, where k
q.i.th
= 0.000833, 0.001667,
and 0.00667 s
1
for the respective bins, S
q.i.th
= 1.0 for all three bins, c
q.i.th
= 2, 5,
and 0 g m
3
for the respective bins, C
q.th
= 3 g m
3
, and C
q.s.i.th
= 1 g m
3
. The
table shows concentrations after 4 h predicted by the APC scheme when each of ve time step
sizes was used. Regardless of the time step size, the APC solution was stable and mass-conserving.
The concentration from this equation cannot fall below zero, but can increase
above the total mass of the species in the system. In such cases, gas concentration
is limited by C
q.t
= min(C
q.t
. C
tot
). This value is substituted into (16.69) to give the
nal aerosol concentration. Since (16.69) can result in a negative concentration or
a concentration above the maximum, two more limits are placed sequentially after
(16.69) has been solved among all size bins. The rst is c
q.i.t
= max(c
q.i.t
. 0) and
the second is
c
q.i.t
=
C
q.th
C
q.t
+
N
B
i =1
max(c
q.i.th
c
q.i.t
. 0)
N
B
i =1
[max(c
q.i.t
c
q.i.th
. 0)]
(c
q.i.t
c
q.i.th
) (16.72)
where the c
q.i.t
values on the right side of the equation are determined after the rst
limit has been solved over all size bins. The APC scheme is unconditionally stable,
since all nal concentrations are bounded between 0 and C
tot
, regardless of the time
step. Table 16.4 demonstrates the unconditional stability and mass conservation
of the scheme.
Figure 16.2 compares an APC scheme prediction with an exact solution for a
case of cloud drop growth from aerosol particles. The gure shows that the APC
scheme matched the exact solution almost exactly for this set of conditions. The
APC scheme can be extended to treat any number of aerosol size distributions
simultaneously, as described in Jacobson (2002).
16.6 SOLVING HOMOGENEOUS NUCLEATION WITH
CONDENSATION
Homogeneous nucleation competes with condensation for a limited amount of
available vapor. If the numerical solution for condensation is operator-split from
that for homogeneous nucleation, the process solved rst may incorrectly consume
more than its share of vapor unless the operator-splitting time interval is short
545
10
1
10
1
10
3
10
5
10
7
10
9
0.1 1 10 100
d
v

(
m
3

c
m
3
)

/

d

l
o
g
1
0
D
p
p
, m)
Initial
Final
Figure 16.2 Comparison of APC solution,

when h = 10 s, with an exact solution of con-
densational growth over 16 size bins when the
moving-center size structure was used. The two
solutions lie almost exactly on top of each other,
thus are indistinguishable. At the start of the
simulation, the relative humidity was increased
from 90 to 100.002 percent and replenished
every 10 s back to 100.002 percent for ve
minutes.
(1 s). Solving homogeneous nucleation simultaneously with condensation elimi-
nates this problem.
Condensation can be solved simultaneously with homogeneous nucleation with
the APC scheme by treating the mass transfer rate (s
1
) to the rst size bin as the
sum of a condensation and a homogeneous nucleation mass transfer rate:
k
q.1.th
= k
q.cond.1.th
+k
q.hom.1.th
(16.73)
where the condensation mass transfer rate is the rate from (16.64), and the homo-
geneous nucleation mass transfer rate is
k
q.hom.1.th
=

q
1
m
q
J
hom.q
C
q.th
S
q.1.th
C
q.s.1.th
(16.74)
In this equation, J
hom.q
is the homogeneous nucleation rate (new particles per
cubic centimeter per second) from (14.23),
q
3
) and
m
q
is the molecular weight (g mol
1
) of a newly nucleated particle,
1
is the vol-
ume (cm
3
) of a single nucleated particle, and the denominator on the right side
is the maximum gas available for homogeneous nucleation at the beginning of a
time step (mol cm
3
). Equation (16.74) is nonzero only if J
hom.q
> 0 and C
q.th
>
S
q.1.th
C
q.s.1.th
. After gas and size-resolved aerosol mole concentrations are solved
with (16.67)(16.72), the newnumber concentration of particles in the smallest size
bin is
n
1.t
= n
1.th
+MAX
(c
q.1.t
c
q.1.th
)
m
q
1
k
q.hom.1.th
k
q.1.th
. 0
(16.75)
546
16.7 Effects of condensation on coagulation
10
2
10
4
10
6
10
8
10
10
0.001 0.01 0.1 1
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0

D
p
p
, m)
Figure 16.3 Number concentration distribution
initially, , and after 8 seconds, ----, of simul-
taneous nucleation and condensation of sulfuric
acidwater with the APC scheme. The initial gas
concentration of sulfuric acid was 15 g m
3
, the
relative humidity was 90 percent, and the temper-
ature was 300 K.
(particles cm
3
). Figure 16.3 shows results from an 8-s simulation of simultaneous
homogeneous nucleation and condensation when a limited amount of gas was
available for both. The gure suggests that, not only did homogeneous nucleation
produce many new particles in this case, but many of the nucleated particles and
the pre-existing particles grew slightly by condensation. In a competition for vapor
between homogeneous nucleation and condensation, the relative importance of
condensation increases with an increasing number of background particles.
16.7 EFFECTS OF CONDENSATION ON COAGULATION
Condensation increases particle volume, and coagulation reduces particle number.
Because both processes occur simultaneously, it is difcult to estimate the effect of
each experimentally. A model can separate the relative effects. Results from one
such simulation are shown in Fig. 16.4. Simulations of coagulation alone, coagu-
lation coupled with growth, and growth alone were performed for an urban size
distribution consisting of four lognormal modes. Two modes had geometric-mean
volume diameters near 0.2 and 0.50.7 m, respectively (Hering and Friedlander
1982; John et al. 1989). Simulations were performed over an 8-h period. The only
growth process considered was sulfuric acid condensation.
The gures show that coagulation alone decreased the number concentration of
particles smaller than 0.2 m in diameter and slightly increased the number con-
centration of particles near 0.3 min diameter. Growth alone caused the initial size
distribution to shift to the right. The shift of particles to larger diameters caused the
volume concentration of those particles to increase as well, as shown in Fig. 16.4(b).
When coagulation was combined with growth, the number concentration of par-
ticles smaller than 0.2 m in diameter decreased, and the number concentration
547
10
1
10
1
10
3
10
5
0.01 0.1 1 10
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0

D
p
p
, m)
Init.
Coag. only
Growth only
Growth
+ coag.
(a)
10
1
10
0
10
1
10
2
10
3
0.01 0.1 1 10
d
v

(
m
3

c
m
3
)

/

d

l
o
g
1
0

D
p
p
, m)
Initial
Coag. only
Growth
only
Growth + coag.
(b)
Figure 16.4 Change in aerosol (a) number con-
centration and (b) volume concentration after
8 h when coagulation alone, growth alone, and
growth coupled to coagulation were considered.
Only H
2
SO
4
(g) condensed. Initial H
2
SO
4
(g),
50 g m
3
. The initial particle number concen-
tration was 45 070 particles cm
3
. T = 298 K.
From Jacobson (1997a).
of particles between 0.2 and 0.5 m in diameter increased. In sum, growth plus
coagulation pushed particles to slightly larger sizes than did growth alone or coag-
ulation alone.
Figure 16.5 shows the growth-only and growth plus coagulation solutions from
Fig. 16.4(a) when the moving-center and full-moving size structures (Chapter 13)
were used. Since the full-moving structure is nondiffusive during growth, the gure
suggests that the moving-center size structure was also relatively nondiffusive.
16.8 ICE CRYSTAL GROWTH
When the temperature drops below 0

C, water vapor may deposit directly onto
aerosol particles to form ice crystals. Whereas most aerosol particle types can
serve as CCN, only a few can serve as ice deposition nuclei (IDN), which are
aerosol particles that water vapor can potentially heterogeneously nucleate on
548
16.8 Ice crystal growth
10
1
10
0
10
1
10
2
10
3
10
4
10
5
10
6
0.01 0.1 1 10
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0
D
p
p
, m)
Growth
+ coag.
(MC)
Growth
+ coag.
(FM)
Growth
only
(MC)
Growth only (FM)
Figure 16.5 Comparison of growth-only and
coagulation plus growth results from Fig.
16.4(a), obtained with the full-moving (FM)
size structure, with those obtained with the
moving-center (MC) size structure.
to form ice crystals. Whereas the concentration of activated CCN can approach
10
3
cm
3
(10
6
L
1
), the concentration of activated IDN is on the order of 0.5
25 L
1
. Common IDN include large hydrophobic particles, soil dust particles, a
small fraction of organic matter, certain viruses and bacteria, and a small fraction
of sea salt. Soot is not a good IDN, except below 15

C and at high supersatu-
rations. Some pollutants deactivate ice nuclei (Pruppacher and Klett 1997).
The rate of mass growth of a single ice crystal, obtained by modifying (16.13),
is
dm
i
dt
=
4
i
D
v.i
( p
q
p
v.I.i
)
D
v.i
L
s
p
v.I.i
a.i
T
L
s
R
v
T
1
+ R
v
T
(16.76)
where
i
is the crystal electrical capacitance (cm), p
v.I.i
is the effective saturation
vapor pressure of water over ice, and L
s
is the latent heat of sublimation (J g
1
).
The electrical capacitance, analogous to the radius in the liquid drop growth
equation, is a function of ice crystal shape only. Ice crystal shape changes with
temperature, relative humidity, and crystal size. The electrical capacitance is indi-
rectly a function of these variables. Parameterizations of electrical capacitance as
a function of crystal shape include
i
=
a
c.i
2 sphere
a
c.i
e
c.i
ln[(1 +e
c.i
)a
c.i
b
c.i
] prolate spheroid
a
c.i
e
c.i
sin
1
e
c.i
oblate spheroid
a
c.i
ln
4a
2
c.i
b
2
c.i
needle
a
c.i
e
c.i
ln[(1 +e
c.i
)(1 e
c.i
)] column
a
c.i
e
c.i
(2 sin
1
e
c.i
) hexagonal plate
a
c.i
thin plate
(16.77)
549
(e.g., Harrington et al. 1995), where a
c.i
and b
c.i
are the lengths of the major
and minor semiaxes (cm), respectively, and e
c.i
=
1 b
2
c.i
a
2
c.i
12
. The major and
minor semiaxes are the maximum and minimum dimensions, respectively, across
a crystal. A prolate spheroid is elongated at the poles and thin at the equator. An
oblate spheroid is attened at the poles and bulging at the equator. The electrical
capacitance of a sphere is its radius.
The effective saturation vapor pressure in (16.76) is p
v.I.i
= S
I.i
p
v.I
, where p
v.I
is the saturation vapor pressure over a at ice surface, given in (2.65), and S
I.i
is
the equilibrium saturation ratio over ice, analogous to (16.46). Equation (16.76)
requires an effective molecular diffusion coefcient and thermal conductivity, given
by (16.18) and (16.26), respectively. The ventilation factor used in these equations
must be modied for ice crystals. An empirical expression for the ventilation factors
of a falling oblate spheroid ice crystal is
F
q.I.i
. F
h.I.i
=
1 +0.14x
2
x - 1.0
0.86 +0.28x x 1.0
(16.78)
where x = x
q.i
for vapor and x = x
h.i
for energy (Pruppacher and Klett 1997), given
in (16.24) and (16.31), respectively. Equation (16.76) can be solved by integration
or with the APC scheme. A solution to growth of liquid drops and ice crystals
simultaneously in a cloud is given in Section 18.8.1.
16.9 SUMMARY
In this chapter, the growth equation for a single, homogeneous, spherical liquid
drop was derived. The equation assumed that, during condensation, a gas diffuses
to and energy conducts away from a particle surface. The reverse occurs during
evaporation. The equation was expanded to a population of particles and to par-
ticles with multiple components. Two numerical solutions to growth equations
an exact numerical solution and a semianalytical solution were discussed. The
growth equations were used to test the importance of condensation relative to
coagulation. The conclusion was that growth plus coagulation moved particles to
larger sizes than did growth alone or coagulation alone. Amethod of solving homo-
geneous nucleation simultaneously with condensation was also given. Finally, an
ice crystal-depositional growth equation was described. In the next chapter, the
discussion of growth is extended to particles in which trace gases dissolve.
16.10 PROBLEMS
16.1 Calculate the diffusion coefcient of ammonia gas in air assuming the air
temperature is 288 K, the air pressure is 1013 hPa, and the air is dry.
16.2 Calculate the diffusion-coefcient correction for collision geometry and
sticking probability for ammonia gas interacting with a particle of (a) 0.1
m in diameter and (b) 10 m in diameter when the air temperature is
288 K, the air pressure is 1013 hPa, and the accommodation coefcient is 0.1.
550
Assume the air is dry. Explain reasons for the difference in results between
the two diameters.
16.3 Calculate the ventilation factor for water vapor interacting with liquid drops
of the following radii: (a) 10 m, (b) 100 m, (c) 1000m. Assume the
temperature is 298 K, the air pressure is 1010 hPa, and the fall speed of
the liquid drop is approximately V
f.i
2r
2
i

p
9
a
, where
p
= drop density
(1 g cm
3
). Ignore water vapor effects on air density.
16.4 Calculate the critical radius and critical supersaturation when a 0.1-m
radius particle contains (a) 0.05 mol kg
1
of an organic compound of molecu-
lar weight 120 g mol
1
, and(b) 0.05 mol kg
1
of sodiumchloride of molecular
weight 58.44 g mol
1
. Assume the organic compound does not dissociate,
sodiumchloride dissociates completely, and T =298.15 K. Which of the two
solutes enhances CCN activation the most? Explain.
16.5 (a) Solve (16.67) and (16.68) simultaneously (analytically) for c
q.i.t
and C
q.t
.
Assume one size bin is present and all variables are constant, except c
q.i.t
and C
q.t
. (Hint: C
q.t
+c
q.i.t
= C
q.th
+c
q.i.th
, since i = 1.)
(b) Discretize the time derivative in the equations from part (a) in rst-order,
backward-difference form, and write out the right side of both equations
implicitly. Solve the resulting equations for c
q.i.t
and C
q.t
.
16.6 Can condensation of a species occur on particles of some sizes while evapo-
ration occurs on particles of other sizes? If so, under what conditions would
this occur?
16.7 In Fig. 16.1, particles above what radius can grow to cloud-sized drops?
Under what condition can all particles grow larger than the critical radius?
Is this condition likely to occur in the atmosphere? Why or why not?
16.8 For what temperature range is (16.76) valid? Can (16.13) be used to simulate
the growth of supercooled liquid water?
16.9 Write a computer script to calculate the equilibrium saturation ratio versus
diameter and temperature for water vapor over a pure liquid water drop from
(16.46). Ignore the solute and radiative heating effects. Calculate the Kelvin
term for d
i
=0.001 to 10m, when T =298 K, and plot the results. Calculate
the term for d
i
= 0.01 m when T varies from 233 K to 315 K. Discuss the
results.
16.10 (a) Write a computer script to calculate
a
,
a
,
a
, D
q
,
q.i
,
q
,
q
, Kn
q.i
, N
Re.i
,
N
Sc.q
, F
q.i
, and D
q.i
assuming that water vapor condenses onto a particle
of radius r
i
= 0.2 m. Assume also that T = 298 K and p
a
= 1013 hPa.
For the fall speed required in the Reynolds-number calculation, assume
V
f.i
= 2r
2
i

p
g(9
a
). Use additional data from the text, where necessary.
Assume the air is dry.
(b) Using information from Problem 16.10(a), calculate
a
,
h.i
, :
a
,
h
, Kn
h.i
,
N
Pr
, F
h.i
, and
d.i
with the computer script.
(c) Using information from Problem 16.10 (a) and (b), calculate L
e
, p
q.s
, S
q.i
,
p
q.s.i
, C
q.s.i
, and D
eff
q.i
withthe computer script. Assume that only curvature
affects S
q.i
.
551
(d) Repeat Problem 16.10 (a), (b), and (c) for T = 273 K. Comment on the
results.
(e) Repeat Problem 16.10 (a), (b), and (c) for r
i
= 1.0 m. Comment on the
results.
16.11 Write a computer script to solve (16.67) and (16.68) with the APC scheme.
Assume three size bins are present (N
B
=3). Also assume k
q.1.th
=0.000 833
s
1
, k
q.2.th
=0.001 667 s
1
, k
q.3.th
=0.006 67 s
1
, c
q.i.th
=0 for all i, S
q.i.th
=
1 for all i, C
q.th
= 10, C
q.s.i.th
= 2, and concentration units are g m
3
.
Plot the mass concentration of particles in each size bin versus time for a
50-min simulation period, using h = 1 min. Compare the result with an
explicit solution to (16.67) and (16.68), when h = 0.01 s. Is this simulation
representative of growth under atmospheric conditions? Why or why not?
552
17
Chemical equilibrium and dissolution processes
T
he change in size and composition of an aerosol depends on several processes,
including nucleation, emission, coagulation, condensation, ice deposition, dis-
solution, heterogeneous reaction, molecular dissociation, solid precipitation, irre-
versible chemistry, interactions with clouds, sedimentation, dry deposition, and
advection. Many of these processes are tightly coupled. Condensation and ice
deposition at the gasparticle interface were discussed in Chapter 16, interactions of
aerosol particles with clouds are discussed in Chapter 18, and irreversible reactions
are examined in Chapter 19. In this chapter, dissolutional growth, molecular dis-
sociation, solid precipitation, and heterogeneous chemical reaction are described.
These four processes can be simulated with reversible (equilibrium) chemical reac-
tions. Dissolution can also be simulated as a nonequilibriumprocess. In the sections
that follow, terminology, equilibrium equations, activity-coefcient equations, the
water equation, and methods of solving equilibriumand nonequilibriumequations
are given.
17.1 DEFINITIONS
An important process that takes place at gasparticle interfaces is dissolution.
Dissolution is the process by which a gas, suspended over an aerosol particle sur-
face, diffuses to and dissolves in a liquid on the surface. The liquid in which a gas
dissolves is a solvent. A solvent makes up the bulk of a solution, and in aerosol
and hydrometeor particles, liquid water is most often the solvent. In some cases,
such as when sulfuric acidwater particles form, the concentration of sulfuric acid
exceeds that of liquid water, and sulfuric acid is the solvent. In this chapter, liquid
water is assumed to be the solvent.
Any gas, liquid, or solid that dissolves in solution is a solute. One or more solutes
plus the solvent make up a solution, which is a homogeneous mixture of substances
that can be separated into individual components upon a change of state (e.g.,
freezing). A solution may contain many solutes. Suspended material (e.g., solids)
may also be mixed throughout a solution. Such material is not considered part of
the solution.
The ability of a gas to dissolve in water depends on the solubility of the gas, which
is the maximum amount of a gas that can dissolve in a given amount of solvent
at a given temperature. Solutions usually contain solute other than the dissolved
553
gas. The solubility of a gas depends on the quantity of other solutes, since such
solutes affect the thermodynamic activity of the dissolved gas in solution. (Ther-
modynamic activity is discussed shortly.) If water is saturated with a dissolved gas,
and if the solubility of the gas changes due to a change in composition or temper-
ature of the solution, the dissolved gas can evaporate from the solution to the gas
phase.
In a solution, dissolved molecules (e.g., HNO
3
, HCl, H
2
SO
4
) may dissociate
(break into simpler components, namely ions). Positive ions, such as H
+
, Na
+
, K
+
,
Ca
2+
, and Mg
2+
, are cations. Negative ions, such as OH
, Cl
, NO
3
, HSO
4
,
SO
4
2
, HCO
3
, and CO
3
2
, are anions. Dissociation is reversible, meaning that
ions can reform a dissolved molecule. Cations and anions may combine to form
solids (crystallize), which precipitate out of solution (e.g., they are no longer part
of the solution). Precipitation is the formation of an insoluble solid compound due
to the buildup in concentration of dissolved ions in a solution. Solids may also
dissociate back to ions, which reenter the solution.
A chemical process that occurs at the surface of a liquid or solid particle is
heterogeneous reaction (Sections 11.1.4 and 11.3.6). A heterogeneous reaction
may occur when a gas molecule collides with another molecule adsorbed to a solid
particle surface or when a gas molecule adsorbs to a surface, diffuses on the surface,
then reacts with another adsorbed molecule on the surface.
Gas dissolution, molecular dissociation, solid precipitation, and heterogeneous
reaction can be described with reversible chemical reactions, also called thermody-
namic equilibriumreactions. The rates of these reactions in the forward and reverse
directions are usually fast. In this chapter, equilibrium reactions and the coupling
of equilibrium reactions to nonequilibrium transfer of gases between gases and
particles, are discussed.
17.2 EQUILIBRIUM REACTIONS
In this section, equilibrium equations are discussed, and temperature-dependent
equilibrium coefcients are derived. An equilibrium reaction describes a reversible
chemical process among solids, liquids, ions, and/or gases. The change in the num-
ber of moles dn
i
of each reactant i = A, B, D, E, etc. during such a process is
described by
dn
D
D+dn
E
E +

dn
A
A +dn
B
B + (17.1)
Each equilibrium reaction must conserve mass. Thus,
i
k
i
(dn
i
)m
i
= 0 (17.2)
where m
i
is the molecular weight of species i, k
i
= +1 for products, and k
i
= 1
for reactants. Dividing each dn
i
by the smallest dn
i
among all species in (17.1)
gives a set of dimensionless stoichiometric coefcients
i
that can be substituted
554
17.2 Equilibrium reactions
into (17.1) to yield
D
D+
E
E +

A
A +
B
B + (17.3)
which is the general form of an equilibrium reaction. In this subsection, several
types of equilibrium reactions are discussed.
17.2.1 Dissolution reactions
Dissolution occurs when a gas dissolves in a liquid. The gasliquid interface may
be at the surface of the ocean, of an aerosol particle, or of a cloud drop. Dissolution
is often (but not always) a reversible process that can be described by reactions of
the form
AB(g)

AB(aq) (17.4)
where (g) indicates a gas, (aq) indicates dissolved in solution, and the stoichiometric
coefcients are unity. In (17.4), the gas and dissolved (solution) phases of species
AB are assumed to be in equilibrium with each other at the gassolution interface;
thus, the number of molecules of AB transferring from the gas to the solution
equals the number of molecules transferring in the reverse direction. Examples of
dissolution reactions include
H
2
SO
4
(g)

H
2
SO
4
(aq)
HNO
3
(g)

HNO
3
(aq)
HCl(g)

HCl(aq) (17.5)
CO
2
(g)

CO
2
(aq)
NH
3
(g)

NH
3
(aq)
17.2.2 Dissociation reactions
Once in solution, dissolved gases often dissociate reversibly into ions. Substances
that undergo partial or complete dissociation in solution are electrolytes. The degree
of dissociation of an electrolyte depends on the acidity of solution, the strength of
the electrolyte, and the concentrations of other ions in solution.
The acidity of a solution is a measure of the concentration of hydrogen ions
(protons or H
+
ions) in solution. Acidity is measured in terms of pH, where
pH = log
10
[H
+
] (17.6)
Here, [H
+
] is the molarity of H
+
(moles of H
+
per liter of solution). The pH scale,
shown in Fig. 17.1 for a limited range, varies from less than 0 (lots of H
+
and very
acidic) to greater than 14 (very little H
+
and very basic or alkaline). Neutral pH,
555
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
Natural
rainwater
(55.6)
Distilled
water
(7.0)
pH
Seawater
(7.88.3)
Battery
acid
(1.0)
Acid
rain, fog
(25.6)
More acidic
More basic or alkaline
Lemon
juice
(2.2)
Vinegar
CH
3
COOH(aq)
(2.8)
Apples
(3.1)
Milk
(6.6)
Baking
soda
NaHCO
3
(aq)
(8.2)
Ammonium
hydroxide
NH
4
OH(aq)
(11.1)
Lye
NaOH(aq)
(13.0)
Slaked lime
Ca(OH)
2
(aq)
(12.4)
Figure 17.1 Diagram of the pH scale and the pH level of selected solutions.
the pH of distilled water, is 7.0. At this pH, the molarity of H
+
is 10
7
mol L
1
.
A pH of 4 means that the molarity of H
+
is 10
4
mol L
1
. Thus, water at a pH of
4 is 1000 times more acidic (contains 1000 times more H
+
ions) than is water at
a pH of 7.
The more acidic a solution, the higher the molarity of H
+
, and the lower the pH.
Hydrogen ions in solution are donated by acids that dissolve. In the atmosphere,
such acids are primarily sulfuric acid, nitric acid, hydrochloric acid, and carbonic
acid.
The ability of an acid to dissociate into protons and anions varies among acids.
At low pH (1), H
2
SO
4
(aq), HNO
3
(aq), and HCl(aq) dissociate readily, whereas
H
2
CO
3
(aq) does not. The former acids are strong acids, and the latter is a weak
acid. Since all acids are electrolytes, a strong acid is a strong electrolyte (i.e., it
dissociates signicantly) and a weak acid is a weak electrolyte.
Sulfuric acid is a strong acid and electrolyte. At pH>3, it dissociates to another
strong acid, the bisulfate ion, which, at pH > +2, dissociates to the sulfate ion.
These two reactions are, respectively,
H
2
SO
4
(aq)

H
+
+HSO

4
HSO

4

H
+
+SO
2
4
(17.7)
Nitric acid is also a strong acid and dissociates signicantly to the nitrate ion
above a pH of 1 by
HNO
3
(aq)

H
+
+NO

3
(17.8)
556
17.2 Equilibrium reactions
Hydrochloric acid is an extremely strong acid because it almost always dissoci-
ates completely to the chloride ion above a pH of 6 by
HCl(aq)

H
+
+Cl
(17.9)
Carbonic acid, which originates from dissolved carbon dioxide, is a weak acid.
It dissociates to the bicarbonate ion above a pH of only +6. The bicarbonate ion
further dissociates to the carbonate ion above a pH of +10. Carbonic acid and
bicarbonate ion dissociation reactions are
CO
2
(aq) +H
2
O(aq)

H
2
CO
3
(aq)

H
+
+HCO

3
(17.10)
HCO

3

H
+
+CO
2
3
respectively. In the background air, the mixing ratio of CO
2
(g) is about 375 ppmv.
A fraction of this CO
2
(g) always dissolves in rainwater. Thus, rainwater, even
in the cleanest environment on Earth, is naturally acidic due to the presence of
background carbonic acid in it. The pHof rainwater affected by only carbonic acid
is about 5.6, indicating that its hydrogen-ion molarity is 25 times that of distilled
water. Other natural acids may reduce the pH of natural rainwater slightly below
5.6. Anthropogenic acids may reduce the pH of rainwater to 2.5 (Fig. 17.1).
Whereas acids provide hydrogen ions, bases provide hydroxide ions (OH
).
Such ions react with hydrogen ions to form neutral water via
H
2
O(aq)

H
+
+OH
(17.11)
Thus, the addition of a base removes hydrogen ions, increasing the pH. An impor-
tant basic substance in aerosol particles is ammonia. In solution, dissolved ammonia
reacts with water to form the ammonium ion and the hydroxide ion by
NH
3
(aq) +H
2
O(aq)

NH
+
4
+OH
(17.12)
Since some strong electrolytes, such as hydrochloric acid and nitric acid, disso-
ciate nearly completely in atmospheric particles, the undissociated forms of these
species are sometimes ignored. Instead, gasion equilibrium reactions, such as
HCl(g)

H
+
+Cl
(17.13)
HNO
3
(g)

H
+
+NO

3
(17.14)
replace the combination of the dissolution and dissociation reactions.
17.2.3 Solid precipitation reactions
At high enough concentrations, ions may precipitate from solution to form solid
electrolytes that are in equilibrium with the ions that formed them in solution. If
ion concentrations decrease (e.g., if the liquid-water content of a solution increases,
causing dilution), a solid electrolyte may also dissociate back to ions. These pro-
cesses are represented by reversible solidion equilibrium reactions. Some such
557
reactions for ammonium-containing electrolytes include
NH
4
Cl(s)

NH
+
4
+Cl
NH
4
NO
3
(s)

NH
+
4
+NO

3
(NH
4
)
2
SO
4
(s)

2NH
+
4
+SO
2
4
(17.15)
Others, for sodium-containing electrolytes, include
NaCl(s)

Na
+
+Cl
NaNO
3
(s)

Na
+
+NO

3
(Na)
2
SO
4
(s)

2Na
+
+SO
2
4
(17.16)
17.2.4 Gassolid heterogeneous reactions
Gassolid heterogeneous reactions may also be represented by reversible reactions.
For example, suppose nitric acid gas collides with and adsorbs to a particle surface,
and an ammonia gas molecule collides with and reacts with the adsorbed nitric acid.
This reaction may result in the formation of solid ammoniumnitrate (NH
4
NO
3
(s)),
which is adsorbed to the surface. The equilibriumreaction representing this process
and a similar one for the formation of solid ammonium chloride (NH
4
Cl(s)) is
NH
4
NO
3
(s)

NH
3
(g) +HNO
3
(g)
NH
4
Cl(s)

NH
3
(g) +HCl(g) (17.17)
In sum, equilibrium relationships can describe gasliquid, gasion, gassolid,
liquidliquid, liquidion, liquidsolid, ionion, and ionsolid reversible reactions.
Appendix Table B.7 lists major equilibrium reactions of atmospheric interest.
17.3 EQUILIBRIUM RELATION AND COEFFICIENTS
The relative quantities of reactants and products in an equilibrium equation are
determined from the equilibrium relation,
i
{a
i
}
k
i
i
=
{A}
A
{B}
B
. . .
{D}
D
{E}
E
. . .
= K
eq
(T) (17.18)
where K
eq
(T) is a temperature-dependent equilibrium coefcient, {a
i
} is the ther-
modynamic activity of species i, {A}, etc., are individual thermodynamic activities,
k
i
= +1 for products, and k
i
= 1 for reactants. Below, thermodynamic activities
and a derivation of the equilibrium relation are described.
17.3.1 Thermodynamic activities
An activity is a relative quantity of a substance and is determined differently for
each phase state. The activity of a gas over a particle surface is its saturation vapor
558
17.3 Equilibrium relation and coefficients
pressure (atm). Thus,
{A(g)} = p
A.s
(17.19)
The activity of an ion in solution or of an undissociated electrolyte is its molality
m
A
(m, moles of solute per kilogramof solvent) multiplied by its activity coefcient
(dimensionless). Thus,
{A
+
} = m
A
+
A
+ {A(aq)} = m
A
A
(17.20)
respectively. A solute activity coefcient represents the deviation from ideal behav-
ior of a solute in solution. It is a dimensionless parameter by which the molality
of a species in solution is multiplied to give the species thermodynamic activity.
In an ideal, innitely dilute solution, the solute activity coefcient is unity. In a
nonideal, concentrated solution, solute activity coefcients may exceed or be less
than unity. Debye and H uckel showed that, in sufciently dilute solutions, where
ions are far apart, the deviation of molality from thermodynamic activity is caused
by Coulombic (electric) forces of attraction and repulsion. At high concentrations,
ions are close together, and ionion interactions affect solute activity coefcients
more than do Coulombic forces.
The activity of liquid water in a particle is the ambient relative humidity. Thus,
{H
2
O(aq)} = a
w
=
p
v
p
v.s
= f
r
(17.21)
where a
w
denotes the activity of water, p
v
is the partial pressure of water vapor, p
v.s
is the saturation vapor pressure of water over a bulk liquid surface, and f
r
is the
relative humidity, expressed as a fraction. For pure liquid water in equilibriumwith
water vapor, p
v
= p
v.s
, and a
w
= 1, which is the water activity in a dilute solution.
When liquid water contains solute, some of the solute replaces liquid water on
the surface of the drop. Water molecules often bond to the solute (Section 17.8),
requiring more water to condense fromthe vapor phase to maintain saturation over
the solution, lowering the partial pressure of water vapor over the drop surface,
decreasing the relative humidity over the surface, and decreasing the water activity.
Solids are not in solution, and their concentrations do not affect molalities or
activity coefcients of solutes in solution. The activity of a pure solid is the mole
fraction of the pure solid in itself. In other words,
{A(s)} = 1 (17.22)
17.3.2 Derivation of the equilibrium relation
The equilibrium relation (17.18) is derived by minimizing the change in Gibbs free
energy of a system. The Gibbs free energy (J) is dened as
G
= H
TS
= U
+ p
a
V TS
(17.23)
where H
= U
+ p
a
V is the enthalpy (J), T is the absolute temperature (K), S
is
the entropy (J K
1
), U
is the internal energy (J), p

a
is the pressure, and V is the
559
volume of a system. The change in Gibbs free energy is a measure of the maximum
amount of useful work that may be obtained from a change in enthalpy or entropy
in the system. It is calculated as
dG
= d(H
TS
) = dU
+ p
a
dV + Vdp
a
T dS
dT (17.24)
where dS
= dQ
T. Thus, the change in entropy of a systemequals the incremental

energy added to or removed from the system divided by the absolute temperature.
The internal energy was introduced in Chapter 2. For a systemin which no changes
in chemical composition occur, it is written as
dU
= dQ
p
a
dV = T dS
p
a
dV (17.25)
For a system in which reversible chemical reactions of the form shown in (17.1)
occur, it is modied to
dU
= T dS
p
a
dV +
i
k
i
(dn
i
) j
i
(17.26)
where j
i
(J mol
1
) is the chemical potential of species i. The chemical potential
is a measure of the intensity of a substance and is a function of temperature and
pressure. Differences in chemical potential between two substances can lead to
chemical reactions between them. Substituting (17.26) into (17.24) gives
dG
= Vdp
a
S
dT +
i
k
i
(dn
i
) j
i
(17.27)
When temperature and pressure are held constant, (17.27) simplies to
dG
i
k
i
(dn
i
) j
i
(17.28)
From this equation, it can be seen that the chemical potential of a substance is
its change in Gibbs free energy per unit change in its number of moles, or its
partial molar free energy, when temperature and pressure are held constant. The
chemical potential can also be dened in terms of a standard chemical potential
and perturbation term. Thus,
j
i
=
i
n
i
T. p
a
= j
o
i
(T) + R
T ln{a
i
} (17.29)
where dG
i
is the change in Gibbs free energy of an individual species at constant
temperature and pressure, R
is the universal gas constant (J mol

1
K
1
), and T
is absolute temperature (K). In the second expression, j
o
i
(J mol
1
), which varies
with temperature, is the standard chemical potential of a substance, and R
T ln{a
i
}
accounts for the deviation in chemical potential from the standard state due to the
activity of the substance.
Thermodynamic equilibriumoccurs when dG
= 0. Setting (17.28) equal to zero

and dividing through by the smallest value of dn
i
among all species in the reaction
560
17.3 Equilibrium relation and coefficients
under consideration give
i
k
i
i
j
i
= 0 (17.30)
where
i
is the dimensionless stoichiometric coefcient. Substituting (17.29) into
(17.30), assuming T = T
0
= 298.15 K for now, dening L
f
G
o
i
= j
o
i
(T
0
) as the
standard molal Gibbs free energy of formation (J mol
1
) of substance i at T = T
0
,
and noting that
i
k
i
i
ln{a
i
} = ln
i
{a
i
}
k
i
i
give
i
k
i
i
j
o
i
(T
0
) + R
T
0
i
k
i
i
ln{a
i
} =
i
k
i
i
L
f
G
o
i
+ R
T
0
ln
i
{a
i
}
k
i
i
= 0
(17.31)
Rearranging (17.31) yields
i
{a
i
}
k
i
i
= exp
1
R
T
0
i
k
i
i
L
f
G
o
i
(17.32)
The right side of (17.32) is the equilibrium coefcient at T = T
0
. Thus,
K
eq
(T
0
) = exp
1
R
T
0
i
k
i
i
L
f
G
o
i
(17.33)
which is independent of species concentration. Substituting (17.33) into (17.32)
when T = T
0
gives the equilibrium-coefcient equation shown in (17.18). Values
of L
f
G
o
i
are given in Appendix Table B.6 for some substances.
17.3.3 Temperature dependence of the equilibrium coefcient
The temperature dependence of the equilibrium coefcient can be estimated by
solving the vant Hoff equation,
dln K
eq
(T)
dT
=
1
R
T
2
i
k
i
i
L
f
H
i
(17.34)
where L
f
H
i
is the molal enthalpy of formation (J mol
1
) of a substance. The vant
Hoff equation is similar in form to the Arrhenius equation given in Chapter 10.
L
f
H
i
can be approximated as a function of temperature with
L
f
H
i
L
f
H
o
i
+c
o
p.i
(T T
0
) (17.35)
where L
f
H
o
i
is the standard molal enthalpy of formation (J mol
1
) (at T
0
=
298.15 K) and c
o
p.i
is the standard molal heat capacity (standard specic heat)
at constant pressure (J mol
1
K
1
). Although c
o
p.i
varies slightly with temperature,
(17.35) assumes that it does not. Combining (17.34) and (17.35) and writing the
result in integral form give
T
T
0
dln K
eq
(T) =
T
T
0
1
R
T
2
i
k
i
L
f
H
o
i
+c
o
p.i
(T T
0
)
dT (17.36)
561
Integrating yields the temperature-dependent expression for the equilibrium coef-
cient,
K
eq
(T) = K
eq
(T
0
) exp
i
k
i
L
f
H
o
i
R
T
0
T
0
T
1
+
c
o
p.i
R
1
T
0
T
+ln
T
0
T
(17.37)
where K
eq
(T
0
) is the equilibrium coefcient at T = T
0
found in (17.33). Values of
L
f
H
o
i
and c
o
p.i
are measured experimentally and given in Appendix Table B.6 for
some substances. Appendix Table B.7 gives temperature-dependent equilibrium-
coefcient expressions for some reactions of the form shown in (17.37) derived
from the data in Table B.6.
Example 17.1
Determine the equilibrium-coefcient expression for Na
2
SO
4
(s)

2Na
+
+
SO
4
2
.
Applying data from Appendix Table B.6 gives
i
k
i
i
L
f
G
o
i
= 1820
J
mol
i
k
i
i
L
f
H
o
i
= 2430
J
mol
i
k
i
i
c
o
p.i
= 328.4
J
mol K
Substituting these expressions into (17.37) and (17.33) yields K
eq
(T) =
0.4799 exp
0.9802
T
0
T
1
+39.497
1 +ln
T
0
T
T
0
T
mol
2
kg
2
17.4 FORMS OF EQUILIBRIUM-COEFFICIENT EQUATIONS
Equation (17.18) expresses an equilibrium reaction in terms of thermodynamic
activities and an equilibrium coefcient. In this section, activity expressions are
substituted for activities to give equilibrium-coefcient equations for several types
of reactions.
17.4.1 Dissolution reactions
For a dissolution reaction, such as HNO
3
(g)

HNO
3
(aq), the equilibrium coef-
cient relates the pressure exerted by a gas at the gasliquid interface to the molality
of the dissolved gas in solution. The equilibrium coefcient expression for HNO
3
dissolution is
{HNO
3
(aq)}
{HNO
3
(g)}
=
m
HNO
3
(aq)
HNO
3
(aq)
p
HNO
3
(g).s
= K
eq
(T)
mol
kg atm
(17.38)
562
17.4 Forms of equilibrium-coefficient equations
where p
HNO
3
(g).s
is the saturation vapor pressure of nitric acid (atm), m
HNO
3
(aq)
is
the molality of nitric acid in solution (mol kg
1
), and
HNO
3
(aq)
is the solute activity
coefcient of dissolved, undissociated nitric acid (dimensionless). In Chapter 2, a
saturation vapor pressure was dened as the partial pressure of a gas when the gas
is in equilibrium with a particle surface at a given temperature. Since HNO
3
(g) is
in equilibrium with the particle surface in this case, its saturation vapor pressure,
not its partial pressure, is used in (17.38).
Over a dilute solution, the partial pressure exerted by a gas is proportional
to the molality of the dissolved gas in solution. This is Henrys law. For a dilute
solution,
HNO
3
(aq)
= 1, and (17.38) obeys Henrys law. In a dilute or concentrated
solution, K
eq
(T) in (17.38) is called a Henrys constant. Henrys constants, like other
equilibrium coefcients, are temperature dependent. Henrys constants for gases
dissolving in liquid water are given in Appendix Table B.7. Species with moderate
or large Henrys constants include NH
3
, HNO
3
, HCl, H
2
SO
4
, H
2
O
2
, and SO
2
.
For these and many other gases

i
k
i
i
L
f
H
o
i
- 0. In such cases, (17.37) predicts
that solubilities and Henrys constants increase with decreasing temperature.
17.4.2 Dissociation and other reactions
In solution, many substances dissociate. The dissociation reaction for nitric acid
has the form HNO
3
(aq)

H
+
+NO

3
. The equilibrium coefcient expression
for this reaction is
{H
+
}{NO

3
}
{HNO
3
(aq)}
=
m
H
+
H
+
m
NO

3
NO

3
m
HNO
3
(aq)
HNO
3
(aq)
=
m
H
+
m
NO

3
2
H
+
.NO

3
m
HNO
3
(aq)
HNO
3
(aq)
= K
eq
(T)
mol
kg
(17.39)
The solute activity coefcients in (17.39) are mixed activity coefcients because
they are determined by considering a mixture of all dissociated and undissoci-
ated electrolytes in solution. The parameters
H
+ and
NO

3
are single-ion mixed
solute activity coefcients, and
H
+
.NO

3
is a mean (geometric-mean) mixed solute
activity coefcient. When HNO
3
(aq), H
+
, and NO

3
are alone in solution,
H
+
and
NO

3
are single-ion binary solute activity coefcients, and
H
+
.NO

3
is a mean
(geometric-mean) binary solute activity coefcient. Activity coefcients for sin-
gle ions are difcult to measure, because single ions cannot be isolated from a
solution. Single-ion solute activity coefcients are generally estimated mathemati-
cally. Mean binary solute activity coefcients are measured in the laboratory. Mean
mixed solute activity coefcients are estimated from mean binary solute activity
coefcient data with a mixing rule.
A geometric mean solute activity coefcient,
, is related to single-ion solute

activity coefcients by
+
+

1(
+
+
)
(17.40)
where
+
and
are the activity coefcients of a cation and anion, respectively, and
+
and
are the stoichiometric coefcients of the cation and anion, respectively.

563
When
+
= 1 and
= 1, the electrolyte is univalent. When

+
> 1 or
> 1, the
electrolyte is multivalent. When
+
=
, the electrolyte is symmetric; otherwise, it

is nonsymmetric. In all cases, a dissociation reaction must satisfy the charge balance
requirement z
+
+
+ z
= 0, where z
+
is the positive charge on the cation and
z
is the negative charge on the anion. Rearranging (17.40) gives
+
+
+
+

(17.41)
which is the form of the mean mixed solute activity coefcient used in (17.39).
Example 17.2
HNO
3
(aq) dissociates by the reaction HNO
3
(aq)

H
+
+NO

3
. Since
+
= 1
and
=1, HNO
3
(aq) is a univalent, symmetric electrolyte. Since z
+
= +1 and
z
= 1, nitric acid dissociation satises z

+
+
+ z
= 0. Na
2
SO
4
(s) dissoci-
ates by the reaction, Na
2
SO
4
(s)

2Na
+
+SO
2
4
. Since
+
= 2 and
= 1,
Na
2
SO
4
(s) is a multivalent, nonsymmetric electrolyte. Since z
+
= +1 and
z
= 2, sodium sulfate dissociation satises z

+
+
+ z
= 0.
Below, equilibriumcoefcient expressions for several additional reactions are given.
In the case of the solid-forming reaction (17.42), the equilibriumcoefcient is called
a solubility product.
Na
2
SO
4
(s)

2Na
+
+SO
2
4
(17.42)
{Na
+
}
2
SO
2
4
{Na
2
SO
4
(s)}
=
m
2
Na
+
2
Na
+
m
SO
2
4
SO
2
4
1.0
= m
2
Na
+m
SO
2
4
3
2Na
+
.SO
2
4
= K
eq
(T)
mol
3
kg
3
HSO

4

H
+
+SO
2
4
(17.43)
{H
+
}
2
SO
2
4
{H
+
}
HSO

4
=
m
2
H
+
2
H
+
m
SO
2
4
SO
2
4
m
H
+
H
+
m
HSO

4
HSO

4
=
m
H
+
m
SO
2
4
3
2H
+
.SO
2
4
m
HSO

4
2
H
+
.HSO

4
= K
eq
(T)
mol
kg
NH
3
(g) +HNO
3
(g)

NH
+
4
+NO

3
(17.44)
NH
+
4

NO

3
{NH
3
(g)}{HNO
3
(g)}
=
m
NH
+
4
NH
+
4
m
NO

3
NO

3
p
NH
3
(g).s
p
HNO
3
(g).s
=
m
NH
+
4
m
NO

3
2
NH
+
4
.NO

3
p
NH
3
(g).s
p
HNO
3
(g).s
= K
eq
(T)
mol
2
kg
2
atm
2
564
17.5 Mean binary solute activity coefficients
NH
3
(aq) +H
2
O(aq)

NH
+
4
+OH
(17.45)
NH
+
4
{OH
}
{NH
3
(aq)}{H
2
O(aq)}
=
m
NH
+
4
NH
+
4
m
OH
OH
m
NH
3
(aq)
NH
3
(aq)
f
r
=
m
NH
+
4
m
OH
2
NH
+
4
.OH
m
NH
3
(aq)
NH
3
(aq)
f
r
= K
eq
(T)
mol
kg
17.5 MEAN BINARY SOLUTE ACTIVITY COEFFICIENTS
Mean binary solute activity coefcients, which are used in mixing rules to determine
mean mixed solute activity coefcients, can be determined from measurement or
theory. Measurements of mean binary solute activity coefcients for several species
have been performed at 298.15 K. Theoretical parameterizations of solute activity
coefcients and their temperature dependences have also been developed.
One parameterization is Pitzers method (Pitzer and Mayorga 1973; Pitzer 1991),
which estimates the mean binary solute activity coefcient of an electrolyte at
298.15 K with
ln
0
12b
= Z
1
Z
2
f
+m
12
2
1
1
+
2
B
12
+m
2
12
2(
1
2
)
32
1
+
2
C
12
(17.46)
where
0
12b
is the mean binary solute activity coefcient of electrolyte 12 (cation 1
plus anion 2) at reference temperature 298.15 K, Z
1
and Z
2
are the absolute values
of the charges of cation 1 and anion 2, respectively; m
12
is the molality of the
electrolyte dissolved in solution, and
1
and
2
are the stoichiometric coefcients
of the dissociated ions. The remaining quantities in (17.46) are
f
= 0.392
I
12
1 +1.2I
12
+
2
1.2
ln(1 +1.2I
12
)
(17.47)
B
12
= 2
(1)
12
+
2
(2)
12
4I
1 e
2I
12
1 +2I
12
2I
(17.48)
where I is the solutions ionic strength (mol kg
1
). Ionic strength is a measure of
interionic effects resulting fromattraction and repulsion among ions. It is calculated
from
I =
1
2
N
C
i =1
m
2i 1
Z
2
2i 1
+
N
A
i =1
m
2i
Z
2
2i
(17.49)
where N
C
and N
A
are the numbers of cations and anions, respectively, in solution.
Odd-numbered subscripts refer to cations, and even-numbered subscripts refer to
anions. In the case of one electrolyte, such as HCl(aq) alone in solution, N
C
=
1 and N
A
=1. The quantities,
(1)
12
,
(2)
12
, and C
12
are empirical parameters
derived from measurements. Pitzer parameters for three electrolytes are shown in
Table 17.1.
565
Table 17.1 Pitzer parameters for three electrolytes
Electrolyte
(1)
12

(2)
12
C

12
HCl 0.17750 0.2945 0.0012
HNO
3
0.1119 0.3206 0.0015
NH
4
NO
3
0.0154 0.112 0.000045
Source: Pilinis and Seinfeld (1987).
3
2
1
0
1
2
3
4
5
0 1 2 3 4 5 6
HNO
3
NH
4
NO
3
HCl
l
n
(
b
i
n
a
r
y

a
c
t
i
v
i
t
y

c
o
e
f
f
i
c
i
e
n
t
)
m
1/2
Figure 17.2 Comparison of mean binary solute
activity coefcients measured by Hamer and Wu
(1972), - - -, with those computed from Pitzers
method, , as a function of the square root of
molality.
Pitzers method predicts mean binary solute activity coefcients at 298.15 K
from physical principles. Its limitation is that the coefcients are typically accurate
only up to about 6 m. Figure 17.2 shows a comparison of mean binary solute
activity coefcients predicted by Pitzers method to those measured by Hamer and
Wu (1972). The measured data are accurate to higher molalities.
Figure 17.2 provides insight into why the chloride ion is forced out of aerosol-
particle solutions in the presence of nitric acid. When sea-spray particles containing
the chloride ion encounter nitric acid gas, the nitric acid enters the particles and
dissociates to the nitrate ion, displacing the chloride to the gas phase. This process
can remove nearly all chloride fromsea-spray particles over the ocean, as discussed
in Section 14.1.1. Upon initial observation of the effective Henrys constants of
HCl(g)

H
+
+ Cl
and HNO
3
(g)

H
+
+ NO
3
, it is not obvious that the

nitrate ion should force the chloride ion out of solution. The effective Henrys-
constant expressions for the two reactions at 298.15 K, obtained from Appendix
Table B.7, are
m
H
+m
Cl
2
H
+
.Cl
p
HCl(g).s
= 1.97 10
6
m
H
+m
NO

3

2
H
+
.NO

3
p
HNO
3
(g).s
= 2.51 10
6
(17.50)
The equilibrium coefcients are comparable for the two reactions, but the binary
solute activity coefcient data in Fig. 17.2 together with (17.50) indicate that, at
566
17.6 Temperature dependence of binary solute activity coefficients
the same molality, m
H
+m
Cl
p
HCl(g)
is much smaller than m
H
+m
NO

3
p
HNO
3
(g)
. The
large binary solute activity coefcient of HCl decreases HCls solubility, causing it
to evaporate. Thus, if NO
3
and Cl
are initially present in solution at the same

time, the nitrate ion is more likely to stay in solution than is the chloride ion.
Whether mean binary solute activity coefcients at 298.15 K are determined
from measurements or theory, they can be parameterized with a polynomial t of
the form
ln
0
12b
= B
0
+ B
1
m
12
12
+ B
2
m
12
+ B
3
m
32
12
+ (17.51)
where B
0
, B
1
, . . . are tting coefcients. Polynomial coefcients for several elec-
trolytes are given in Appendix Table B.9. Polynomial ts are used to simplify and
speed up the use of binary solute activity coefcient calculations in computer pro-
grams.
17.6 TEMPERATURE DEPENDENCE OF BINARY SOLUTE
ACTIVITY COEFFICIENTS
The temperature dependence of mean binary solute activity coefcients can be
derived fromthermodynamic principles. An expression for the binary solute activity
coefcient of electrolyte 12 at temperature T is (Harned and Owen 1958)
ln
12b
(T) = ln
0
12b
+
T
L
(
1
+
2
)R
T
0
L
+m
L
m
+
T
C
(
1
+
2
) R
c
p
+m
c
p
m

o
c
p
(17.52)
where T
0
is the reference temperature (298.15 K), R
is the universal gas constant

(J mol
1
K
1
),
L
is the relative apparent molal enthalpy (J mol
1
) of the electrolyte
at molality m (with subscript 12 omitted),
c
p
is the apparent molal heat capacity
(J mol
1
K
1
) at molality m,
o
c
p
is the apparent molal heat capacity at innite
dilution, and
T
L
=
T
0
T
1 T
C
= 1 +ln
T
0
T
T
0
T
(17.53)
The relative apparent molal enthalpy equals the negative of the heat of dilution.
Appendix Table B.9 lists several sources for heat of dilution and apparent molal
heat capacity experimental data. With these data, polynomials of the form
L
= U
1
m
12
+U
2
m+U
3
m
32
+
(17.54)
c
p
=
0
c
p
+ V
1
m
12
+ V
2
m+ V
3
m
32
+
can be constructed, where U
1....
and V
1....
are polynomial coefcients. The relative
apparent molal enthalpy is dened as
L
=
H
0
H
, where
H
is the apparent
567
10
2
10
1
10
0
10
1
10
2
10
3
10
4
10
5
10
6
0 1 2 3 4 5 6 7 8
201 K
273 K
298 K
328 K
B
i
n
a
r
y

a
c
t
i
v
i
t
y

c
o
e
f
f
i
c
i
e
n
t
m
1/2
H
+
/ HSO
4
2H
+
/ SO
4
2
Figure 17.3 Binary solute activity coefcients of sulfate and bisul-
fate, alone in solution as a function of square root of molality (Jacob-
son et al. 1996b). The curves were obtained by combining equations
from Clegg and Brimblecombe (1995) with Equations (72) and (73)
of Stelson et al. (1984) in a NewtonRaphson iteration. The results
are valid for 040 m total H
2
SO
4
.
molal enthalpy and
0
H
is the apparent molal enthalpy at innite dilution, which
occurs when m = 0. At m = 0,
H
=
0
H
and
L
= 0.
Equations (17.51)(17.54) can be combined to give temperature-dependent
mean-binary-solute-activity-coefcient polynomials of the form
ln
12b
(T) = F
0
+ F
1
m
12
+ F
2
m+ F
3
m
32
+ (17.55)
where
F
0
= B
0
F
j
= B
j
+ G
j
T
L
+ H
j
T
C
for j = 1. . . . (17.56)
G
j
=
0.5( j +2)U
j
(
1
+
2
)R
T
0
H
j
=
0.5( j +2)V
j
(
1
+
2
)R
(17.57)
Appendix Table B.9 lists B, G, and H for 10 electrolytes and maximum molalities
associated with the ts for calculating temperature- and molality-dependent mean
binary solute activity coefcients from (17.55)(17.57). Since polynomial ts wan-
der when the molality increases much beyond the limit of the given data, activity
coefcients from beyond the last valid molality should be used cautiously when
modeling.
Determining temperature-dependent binary solute activity coefcients of bisul-
fate and sulfate is more difcult. Figure 17.3 shows calculated solute activity coef-
cients for these electrolytes as a function of temperature and molality and discusses
how they were determined. Such coefcients can be written in polynomial or tab-
ular form.
17.7 MEAN MIXED SOLUTE ACTIVITY COEFFICIENTS
In aerosol particles, multiple electrolytes, such as dissolved sulfuric acid, nitric acid,
hydrochloric acid, ammonia, and sodiumchloride, often coexist in solution. In such
568
17.7 Mean mixed solute activity coefficients
cases, mean mixed solute activity coefcients are needed. They can be approximated
with an empirical mixing rule that is a function of mean binary solute activity
coefcients and accounts for interactions among ions. Two common mixing rules
are Bromleys method (Bromley 1973) and the KusikMeissner method (Kusik
and Meissner 1978). Bromleys method, discussed here, gives the mean activity
coefcient of electrolyte 12 in a mixture as
log
10
12m
(T) = A
Z
1
Z
2
I
12
m
1 +I
12
m
+
Z
1
Z
2
Z
1
+Z
2
W
1
Z
1
+
W
2
Z
2
(17.58)
where A
is the DebyeH uckel parameter (0.392 at 298 K), Z

1
and Z
2
are the
absolute-value charges of cation 1 and anion 2, respectively, I
m
is the total ionic
strength of the mixture, and
W
1
= Y
21
log
10
12b
(T) + A
Z
1
Z
2
I
12
m
1 +I
12
m
+Y
41
log
10
14b
(T) + A
Z
1
Z
4
I
12
m
1 +I
12
m
+ (17.59)
W
2
= X
12
log
10
12b
(T) + A
Z
1
Z
2
I
12
m
1 +I
12
m
+ X
32
log
10
32b
(T) + A
Z
3
Z
2
I
12
m
1 +I
12
m
+ (17.60)
are functions of all electrolytes in solution. In the last equations,
Y
21
=
Z
1
+Z
2
2
2
m
2.m
I
m
X
12
=
Z
1
+Z
2
2
2
m
1.m
I
m
(17.61)
Similar expressions can be written for X
32
, X
52
, . . . , Y
41
, Y
61
, . . . In these
equations,
12b
(T),
14b
(T),
32b
(T), . . . . are temperature-dependent mean binary
solute activity coefcients, odd-numbered subscripts refer to cations, and even-
numbered subscripts refer to anions. Thus, m
1.m
and m
2.m
are the molalities of a
cation and anion, respectively, in the mixture.
When a mean binary solute activity coefcient is used in (17.59) and (17.60), it
is determined at a binary electrolyte molality m
12.b
that results in the current ionic
strength of the mixture, I
m
. The molality of binary electrolyte 12 is found from
I
m
=
1
2
m
1.b
Z
2
1
+m
2.b
Z
2
2
=
1
2
+
m
12.b
Z
2
1
+
m
12.b
Z
2
2
(17.62)
where m
1.b
=
+
m
12.b
and m
2.b
=
m
12.b
are the molalities of a cation and an
anion, alone in solution, that results in the total ionic strength of the mixture.
Solving (17.62) for m
12.b
gives
m
12.b
=
2I
m
+
Z
2
1
+
Z
2
2
(17.63)
569
The molality of the binary electrolyte in (17.63) is used to determine
12b
(T) from
the polynomial t given in (17.55).
17.8 THE WATER EQUATION
Aerosol particles containing solute often absorb liquid water at relative humidities
less than 100 percent, causing the particles to swell. Solutes attracting water include
inorganic and organic ions, undissociated electrolytes, and nonelectrolytes. Some
inorganic electrolytes that attract water include sulfuric acid, nitric acid, ammo-
nium sulfate, and sodium chloride. Some organic compounds that attract water
include succinic acid, glutaric acid, oxalic acid, and malonic acid.
The bonding of a solute to a solvent, such as liquid water, in solution is called
solvation. When the solvent is liquid water, the bonding is called hydration. The
solute may be a cation, anion, undissociated electrolyte, or a nonelectrolyte (e.g.,
sucrose). During hydration of an anion, the water molecule attaches to the anion
end of the ion dipole via hydrogen bonding. Several water molecules can hydrate
to each ion. During hydration of a cation, the lone pair of electrons on the oxygen
atom of a water molecule bonds to the cation end of the ion dipole.
When the relative humidity is less than 100 percent in the atmosphere, all aerosol
liquid water content is due to hydration. Thus, for example, if a sea-spray drop
emitted fromthe ocean surface did not contain sodium, chloride, and other ions that
hydrate, the drop would evaporate entirely when it entered sub-100-percent relative
humidity air. The quantity of water taken up by hydration depends on the molal-
ity of the solute and the relative humidity. The higher the molality and the higher
the relative humidity, the greater the water uptake. Conversely, when the rela-
tive humidity decreases, the liquid water content decreases as well. Thus, when a
sea-spray drop travels from near the ocean surface, where the relative humidity
is near 100 percent, to the boundary layer, where it may be 80 percent, the liq-
uid water content of the drop decreases. For some solutes, such as sulfuric acid,
hydration can occur down to relative humidities near 0 percent. For other solutes,
such as sodium chloride, water uptake stops when the relative humidity drops
below a certain relative humidity. In all cases, as the relative humidity approaches
100 percent, the rate of water uptake increases superlinearly with increasing rela-
tive humidity.
When a gas, such as sulfuric acid, condenses onto a dry particle surface, the
saturation vapor pressure of water over the liquid sulfuric acid is so low that water
vapor condenses immediately. Once condensed, the liquid water hydrates with the
sulfuric acid, creating a sulfuric acidwater solution. In the solution, other gases
may dissolve and absorb more water from the surrounding air, causing the particle
to swell further. The reduction of the saturation vapor pressure of water over a
particle surface due to the presence of a solute is Raoults law (the solute effect),
discussed in Section 16.2.3.2. The solute effect, through hydration, allows particles
to absorb liquid water when the relative humidity is less than 100 percent.
Above 100 percent relative humidity, water vapor cascades onto particles by
condensation. In such cases, the volume of water added to particles by hydration is
570
17.8 The water equation
Table 17.2 Experimental test of ZSR prediction accuracy for a
mixture of sucrose (species a) and mannitol (species b) at two
water activities
Case m
x.a
m
y.a
m
x.m
m
y.m
m
x.m
m
x.a
+
m
y.m
m
y.a
1 0.7751 0.8197 0.6227 0.1604 0.9990
2 0.9393 1.0046 0.1900 0.8014 1.0000
Source: Stokes and Robinson (1966).
small compared with that added by condensation. Thus, hydration does not affect
water content much when the relative humidity exceeds 100 percent.
A method of estimating aerosol liquid-water content as a function of elec-
trolyte molality and sub-100-percent relative humidity is the ZdanovskiiStokes
Robinson (ZSR) equation (Stokes and Robinson 1966; Clegg et al. 2003; Clegg
and Seinfeld 2004). The equation can be applied to electrolytes or nonelectrolytes.
The simplest form of the equation is
m
x.m
m
x.a
+
m
y.m
m
y.a
= 1 (17.64)
where m
x.a
and m
y.a
are the molalities of solutes x and y each alone in solution at a
given water activity, and m
x.m
and m
y.m
are the molalities of x and y mixed together
in solution at the same water activity. The water activity for atmospheric particles
was dened in (17.21) as the relative humidity. Table 17.2 shows experimental
data validating (17.64).
Equation (17.64) can be generalized for a mixture of any number of solutes with
k
m
k.m
m
k.a
= 1 (17.65)
where m
k,m
is the molality of solute k in a solution containing all solutes at the
ambient relative humidity, and m
k,a
is the molality of solute k alone in solution at
the ambient relative humidity.
Experimental data for water activity as a function of electrolyte molality are
available (e.g., Robinson and Stokes 1955; Pitzer and Mayorga 1973; Cohen et al.
1987a,b; Tang and Munkelwitz 1994; Tang 1997; Peng et al. 2001). Such data can
be tted to polynomial expressions of the form
m
k.a
= Y
0.k
+Y
1.k
a
w
+Y
2.k
a
2
w
+Y
3.k
a
3
w
+ (17.66)
where a
w
is the water activity (relative humidity expressed as a fraction), and the Ys
are polynomial coefcients, listed in Appendix Table B.10 for several electrolytes.
571
0 10 20 30 40 50 60
0
0.2
0.4
0.6
0.8
1
Molality
W
a
t
e
r

a
c
t
i
v
i
t
y
NaNO
3
HNO
3
H
2
SO
4 HCl
0 5 10 15 20 25 30
0.4
0.5
0.6
0.7
0.8
0.9
1
Molality
W
a
t
e
r

a
c
t
i
v
i
t
y
NH
4
NO
3
NH
4
Cl
NaCl
(NH
4
)
2
SO
4
Na
2
SO
4
(a) (b)
Figure 17.4 Water activity as a function of molality for several electrolytes at
298.15 K. The curves were obtained from (17.66) and the coefcients in Appendix
Table B.10.
Figures 17.4 (a) and (b) show molality versus water activity for some of these
electrolytes.
In comparison with the temperature dependence of binary solute activity coef-
cients, the temperature dependence of the water activity coefcients is small under
lower tropospheric conditions. This temperature dependence of the water activity
is
lna
w
(T) = lna
0
w
m
v
m
2
k.a
R
T
L
T
0
L
m
k.a
+ T
C
c
P
m
k.a
(17.67)
(e.g., Harned and Owen 1958; Jacobson et al. 1996b) where m
v
is the molecu-
lar weight of water (0.01802 kg mol
1
here) and the remaining parameters were
dened in (17.52). If the water activity at the reference temperature is extracted
from (17.66) and expressed as
lna
0
w
= A
0
+ A
1
m
12
k.a
+ A
2
m
k.a
+ A
3
m
32
k.a
+ (17.68)
then (17.67), (17.68), and (17.54) can be combined to form the polynomial
lna
w
(T) = A
0
+ A
1
m
12
k.a
+ A
2
m
k.a
+ E
3
m
32
k.a
+ E
4
m
2
k.a
+ (17.69)
where
E
l
= A
l

0.5(l 2)m
v
R
T
L
T
0
U
l2
+ T
C
V
l2
for l = 3. . . . (17.70)
and the Us and Vs are polynomial coefcients from (17.54). Since the A-terms
in (17.69) are not affected by temperature, temperature affects the water activity
polynomial starting with the fourth term, E
3
. In (17.55) temperature affected the
solute activity polynomial starting with the second term. Thus, the temperature
dependence of solute activity coefcients is greater than that of water activity coef-
cients (Jacobson et al. 1996b).
572
17.8 The water equation
Table 17.3 Three sets of hypothetical electrolyte
concentrations that may result when 6 mol m
3
of H
+
,
6 mol m
3
Na
+
, 7 mol m
3
of Cl
, and 5 mol m
3
of NO
3
are combined
Case c
HCl.m
c
HNO
3
.m
c
NaCl.m
c
NaNO
3
.m
1 6 0 1 5
2 4 2 3 3
3 1 5 6 0
Case 1 was obtained from (17.72). Units are micromoles per
cubic meter.
Example 17.3
At high molalities (>10) and ambient temperatures (273310 K), temperature
affects water activity only slightly. For example, at m=16 m, the water activity
of HCl increases from 0.09 to only 0.11 when the temperature increases from
273 K to 310 K. At low molalities, temperature has even less of an effect.
The water equation for mixed aerosol particles can be rewritten from (17.65)
to
c
w
=
1
m
v
N
C
i =1
N
A
j =1
c
i. j.m
m
i. j.a
(17.71)
where c
w
is the aerosol liquid-water content in units of mole concentration (moles of
H
2
O(aq) per cubic centimeter of air), m
v
is the molecular weight of water (0.018 02
kg mol
1
), c
i. j.m
= m
i. j.a
c
w
m
v
is the number of moles of electrolyte pair i, j per cubic
centimeter of air in a solution containing all solutes at the ambient relative humidity,
and m
i. j.a
is the molality (mol kg
1
) of the electrolyte pair alone in solution at the
ambient relative humidity, found from(17.66). The subscript i, j in (17.71) replaces
the subscript k in (17.65).
In a model, the mole concentrations of electrolyte pairs are not usually known,
but those of individual ions are. Thus, ions must be combined into hypothetical
electrolyte pairs for (17.71) to be solved. The easiest way to recombine ions into
hypothetical electrolytes in a mixture is to execute the equations
c
i. j.m
= min
c
i.m
i
.
c
j.m
c
i.m
= c
i.m
i
c
i. j.m
c
j.m
= c
j.m
j
c
i. j.m
(17.72)
in succession, for each undissociated electrolyte pair i, j, where c
i. j.m
is the mole
concentration of undissociated electrolyte i, j (e.g., HNO
3
, HCl), and c
i.m
and
c
j.m
are cation and anion mole concentrations, respectively. Applying (17.72) to
Example 17.4 gives case 1 of Table 17.3.
573
Example 17.4
In a solution containing a mixture of ions, the combination of ions to form
hypothetical electrolytes can be arbitrary for use in the water equation. Sup-
pose a solution contains 6 mol m
3
of H
+
, 6 mol m
3
Na
+
, 7 mol m
3
of
Cl
, and 5 mol m
3
of NO
3
. This set of concentrations is charge conserving

(i.e., the sum, over all ions, of mole concentration multiplied by ion charge
equals zero), which is a requirement for any aqueous system. The four ions
can be combined in many ways to form electrolytes, HNO
3
, HCl, NaNO
3
, and
NaCl. Recombination is limited by the mole-balance constraints
c
H
+
.m
= c
HNO
3
.m
+c
HCl.m
c
Na
+
.m
= c
NaNO
3
.m
+c
NaCl.m
c
Cl
.m
= c
HCl.m
+c
NaCl.m
c
NO
3
.m
= c
HNO
3
.m
+c
NaNO
3
.m
Table 17.3 shows three sets of electrolyte concentrations that satisfy these
constraints.
17.9 SOLID FORMATION AND DELIQUESCENCE
RELATIVE HUMIDITY
Solid electrolytes form by precipitation of ions or liquids from a particle solution
or by chemical reaction of gases on a particle surface. Precipitation is the formation
of an insoluble compound from a solution, and it can be simulated as a reversible
equilibrium process. The equilibrium coefcient of a precipitation reaction is a
solubility product. A solid precipitates from solution when the product of its reac-
tant ion concentrations and mean solute activity coefcient exceeds its solubility
product. For the reaction NH
4
NO
3
(s)

NH
+
4
+NO

3
, precipitation of ammo-
nium nitrate from solution occurs when
m
NH
+
4
m
NO

3
2
NH
+
4
.NO

3
> K
eq
(T) (17.73)
Asolid may also formas a result of heterogeneous reaction of gases on a particle
surface. The solid may form when one gas adsorbs to a surface and the other gas
collides and reacts with the adsorbed gas, or when both gases adsorb to the surface,
diffuse on the surface, and collide. Solid formation by gas interaction with a particle
surface can be simulated with an equilibrium reaction, such as NH
4
NO
3
(s)

NH
3
(g) +HNO
3
(g). In this case, a solid forms when
p
NH
3
(g).s
p
HNO
3
(g).s
> K
eq
(T) (17.74)
Whether a solid forms or dissolves depends on the relative humidity and whether
the relative humidity is increasing or decreasing. If an aerosol particle consists of
an initially solid electrolyte at a given relative humidity, and the relative humid-
ity increases, the electrolyte does not take on liquid water by hydration until the
deliquescence relative humidity (DRH) is reached. Deliquescence is the process by
which an initially dry particle lowers its saturation vapor pressure and takes up
574
17.10 Example equilibrium problem
Table 17.4 DRHs and CRHs of several electrolytes at 298 K
DRH CRH DRH CRH
Electrolyte (percent) (percent) Electrolyte (percent) (percent)
NaCl 75.28
a
47
c
NH
4
HSO
4
40.0
b
-522
b
Na
2
SO
4
84.2
b
5759
b
NH
4
NO
3
61.83
a
2532
d
NaHSO
4
52.0
d
-5
d
(NH
4
)
3
H(SO
4
)
2
69
b
3544
b
NaNO
3
74.5
d
-530
b
KCl 84.26
a
62
c
NH
4
Cl 77.1
a
47
e
Oxalic acid 97.3
f
51.856.7
f
(NH
4
)
2
SO
4
79.97
a
3740
b
Succinic acid 98.8
f
55.259.3
f
a
Robinson and Stokes (1955),
b
Tang and Munkelwitz (1994),
c
Tang (1997),
d
Tang (1996),
e
Cohen et al. (1987a),
f
Peng et al. (2001)
liquid water. At the DRH, water rapidly hydrates with the electrolyte, dissolv-
ing the solid, and increasing the liquid-water content of the particle. Above the
DRH, the solid no longer exists, and the particle takes up liquid water to maintain
equilibrium.
If the particle initially consists of an electrolyte dissolved in water, and the rel-
ative humidity decreases below the DRH, liquid water evaporates, but dissolved
ions in solution do not necessarily precipitate (crystallize) immediately. Instead,
the solution becomes supersaturated with respect to water and remains so until
nucleation of a solid crystal occurs within the solution. The relative humidity at
which nucleation occurs and an initially aqueous electrolyte becomes crystalline
is the crystallization relative humidity (CRH). The CRH is always less than or
equal to the DRH. Table 17.4 shows the DRHs and CRHs of several electrolytes at
298 K.
Some electrolytes, such as NH
3
, HNO
3
, HCl, and H
2
SO
4
do not have a solid
phase at room temperature. These substances, therefore, do not have a DRH or a
CRH.
For an initially dry mixture of two or more electrolytes, the relative humidity
above which the mixture takes on liquid water is the mutual deliquescence relative
humidity (MDRH). Above the MDRH, mixed particles may consist of one or more
solids in equilibrium with the solution phase. The MDRH of a mixture is always
lower than the DRH of any electrolyte within the mixture. (Wexler and Seinfeld
1991; Tang and Munkelwitz 1993).
17.10 EXAMPLE EQUILIBRIUM PROBLEM
Equilibrium equations, activity-coefcient equations, and the water equation are
solved together with other equations in an atmospheric model to determine aerosol
composition and liquid-water content. In this section, a basic equilibrium problem
and the set of equations required to solve it are given.
In this problem, aerosol particles are assumed to contain a solution of water, H
+
,
Cl
, HSO
4
and SO
4
2
, all of which are in equilibrium with each other. HCl(g) is
575
also assumed to be in equilibriumwith the particles. The two equilibriumreactions
describing this situation are
HCl(g)

H
+
+Cl
HSO

4

H
+
+SO
2
4
(17.75)
The equilibrium-coefcient expressions for these reactions are
m
H
+
.eq
m
Cl
.eq
2
H
+
.Cl
.eq
p
HCl.s.eq
= K
eq
(T)
m
H
+
.eq
m
SO
2
4
.eq
3
2H
+
.SO
2
4
.eq
m
HSO

4
.eq
2
H
+
.HSO

4
.eq
= K
eq
(T)
(17.76)
respectively, where the subscript eq indicates that the value is at equilibrium, and
the subscript s indicates a saturation vapor pressure. The activity coefcients are
nonlinear functions of H
+
, Cl
, HSO
4
, and SO
4
2
mole concentrations.
When the reactions in (17.75) are solved, moles of individual atoms (e.g., chlo-
rine, sulfur) must be conserved. The mole balance constraints are
C
HCl(g).eq
+c
Cl
.eq
= C
HCl(g).th
+c
Cl
.th
(17.77)
c
HSO

4
.eq
+c
SO
2
4
.eq
= c
HSO

4
.th
+c
SO
2
4
.th
(17.78)
where Cis the mole concentration of a gas (mol cm
3
), c is the mole concentration of
a particle component, and t h indicates an initial value. Saturation vapor pressures
and molalities are related to mole concentrations by
p
HCl.s.eq
= C
HCl(g).s.eq
R
T m
Cl
.eq
=
c
Cl
.eq
c
w.eq
m
v
(17.79)
respectively. The charge balance equation for this problem requires
c
Cl
.eq
+c
HSO

4
.eq
+2c
SO
2
4
.eq
= c
H
+
.eq
(17.80)
and the ZSR equation giving the water content is
c
w.eq
=
1
m
v
c
H
+
.Cl
.m
m
H
+
.Cl
.a
+
c
H
+
.HSO

4
.m
m
H
+
.HSO

4
.a
+
c
2H
+
.SO
2
4
.m
m
2H
+
.SO
2
4
.a
(17.81)
The molality of a species alone in solution (m
a
) is determined from an empirical
function of the relative humidity, such as that given in (17.66). The quantities
c
H
+
.Cl
.m
, etc., are hypothetical mole concentrations of electrolyte pairs, constrained
by
c
H
+
.eq
= c
H
+
.Cl
.m
+c
H
+
.HSO

4
.m
+2c
2H
+
.SO
2
4
.m
(17.82)
c
Cl
.eq
= c
H
+
.Cl
.m
c
HSO

4
.eq
= c
H
+
.HSO

4
.m
c
SO
2
4
.eq
= c
2H
+
.SO
2
4
.m
This problem requires the solution of two equilibrium equations, two mole
balance equations, a charge balance equation, a water equation, and activity-
coefcient equations. These equations or a variation of them can be solved with an
iterative NewtonRaphson method (e.g., Press et al. 1992), an iterative bisectional
Newton method (e.g., Pilinis and Seinfeld 1987; Kim et al. 1993a,b; Nenes et al.
1998, 1999; Makar et al. 2003), an iterative method that minimizes the Gibbs free
576
17.11 Mass-flux iteration method
energy (Bassett and Seinfeld 1983, 1984; Wexler and Seinfeld 1990, 1991; Wexler
and Clegg 2002), approximation method (e.g., Saxena et al. 1986; Metzger et
al. 2002), or a mole-fraction-based thermodynamic model that parameterizes ion
pair and triplet interactions (Clegg et al. 1997). Most of these methods are mole-
and charge-conserving. Zhang et al. (2000) compare several aerosol equilibrium
models.
17.11 MASS-FLUX ITERATION METHOD
Another method used to solve equilibriumproblems is the mass-ux iteration (MFI)
method (Jacobson 1994; Jacobson et al. 1996b), which relies in part on the tech-
nique of Villars (1959). This method can converge thousands of equilibrium equa-
tions simultaneously, cannot produce negative concentrations, and is mole- and
charge-conserving at all times. The only constraints are that the equilibrium equa-
tions must be mole- and charge-conserving, and the system must start in charge
balance. For example, the equation HNO
3
(aq) = H
+
+NO
3
conserves moles of
H, O, and N and charge. The charge balance constraint allows initial charges to be
distributed among all dissociated ions, but the initial sum, over all species, of charge
multiplied by molality must equal zero. The simplest way to initialize charge is to
set all ion molalities to zero. Initial mass in the systemcan be distributed arbitrarily,
subject to the charge balance constraint. If the total nitrate in the system is known
to be, say, 20 g m
3
, the nitrate can be distributed initially in any proportion
among HNO
3
(aq), NO
3
, NH
4
NO
3
(s), etc.
The MFI method requires the solution of one equilibrium equation at a time by
iteration. A system of equations is solved by iterating all equations many times.
Suppose a systemconsists of a single aerosol size bin and 15 equations representing
the equilibriumchemistry within that bin. At the start, the rst equation is iterated.
When the rst equation converges, the updated and other initial concentrations are
used as inputs into the second equation. This continues until the last equation has
converged. At that point, the rst equation is no longer converged, since concen-
trations used in it have changed. The iteration sequence must be repeated over
all equations several times until the concentrations no longer change upon more
iteration.
Equilibrium among multiple particle size bins and the gas phase is solved in a
similar manner. Suppose a system consists of several size bins, equations per bin,
and gases that equilibrate with dissolved molecules in each size bin. This system
can be solved with equilibrium equations if it is assumed that each gass saturation
vapor pressure over each particle surface equals the gass partial pressure, which
is a single value. In reality, a gass saturation vapor pressure differs over each par-
ticle surface. In order to take variations in saturation vapor pressure over particle
surfaces into account, nonequilibrium gasaerosol transfer equations, discussed in
Section 17.14, must be solved.
Gasparticle equilibrium over multiple size bins can be solved by iterating
each equilibrium equation, including gasparticle equations, starting with the rst
size bin. Updated gas concentrations from the rst bin affect the equilibrium
577
distribution in subsequent bins. After the last size bin has been iterated, the sequence
is repeated in reverse order (to speed convergence), from the last to rst size bin.
The marches back and forth among size bins continue until gas and aerosol con-
centrations do not change upon further iteration.
To demonstrate the MFI method, an example in which two gases equilibrate with
two ions in a single equilibrium equation is discussed. The sample equation has the
form of (17.3), with the gases on the left side of the equation. The rst step in the
solution method is to calculate Q
d
and Q
n
, the smallest ratio of mole concentration
to the stoichiometric coefcient among species appearing in the denominator and
numerator, respectively, of (17.18). Thus,
Q
d
= min
C
D.0
D
.
C
E.0
Q
n
= min
c
A.0
A
.
c
B.0
(17.83)
where C and c are gas- and particle-phase mole concentrations (mol cm
3
), respec-
tively, and the subscript 0 indicates an initial value. If an equilibrium equation
contains a solid, the solids concentration is included in (17.83).
After two parameters, z
1
= 0.5(Q
d
+ Q
n
) and Lx
1
= Q
d
z
1
, where the sub-
script 1 refers to the iteration number, are initialized, the iteration sequence begins
by adding the mass (mole) ux factor Lx, which may be positive or negative,
to each mole concentration in the numerator and subtracting it from each mole
concentration in the denominator of the equilibrium equation. Thus,
c
A.l+1
= c
A.l
+
A
Lx
l
c
B.l+1
= c
B.l
+
B
Lx
l
(17.84)
C
D.l+1
= C
D.l

D
Lx
l
C
E.l+1
= C
E.l

E
Lx
l
Starting with (17.84), iteration numbers are referred to by subscripts l and
l +1. If the equation contains a solid, the change in the solids concentration is cal-
culated with (17.84) (solid, aqueous, and ionic mole concentrations are identied
with a c). Equation (17.84) shows that, during each iteration, moles and charge are
transferred from either reactants to products or vice versa. This transfer continues
until Lx = 0. Thus, the scheme conserves moles and charge each iteration.
Third, a ratio comparing activities to the equilibrium coefcient is calculated
as
F =
m
A
A.l+1
m
B
B.l+1
A
+
B
AB.l+1
p
D
D.l+1
p
E
E.l+1
1
K
eq
(T)
(17.85)
For this calculation, mole concentrations are converted to molalities (for solutes)
or atmospheres (for gases). In the case of solids, the activities are unity and do
not appear in (17.85). Mean mixed solute activity coefcients (e.g..
AB.l+1
) are
updated before each iteration of an equilibrium equation and converge over time.
The liquid-water content c
w
, which also converges, is updated with (17.71) before
each iteration of an equilibrium equation or more sporadically.
578
17.12 Analytical equilibrium iteration method
The fourth step in the procedure is to recalculate z for the next iteration as
z
l+1
= 0.5z
l
. The nal step is to check convergence with the criterion
F =
>1 Lx
l+1
= z
l+1
- 1 Lx
l+1
= +z
l+1
= 1 convergence
(17.86)
At each nonconvergence, Lx is updated, the iteration number is advanced, and
the iteration sequence restarts with (17.84). All molalities eventually converge to
positive numbers.
When a solid forms during iteration of a gassolid or ionsolid equilibrium
reaction, F converges to 1, and the solid concentration is found from (17.84).
When a solid does not form, F cannot converge, and (17.84) correctly predicts no
net change in concentration of any species involved in the reaction.
17.12 ANALYTICAL EQUILIBRIUM ITERATION METHOD
An advantage of the MFI method is that it can be used to solve equilibrium equa-
tions of any form. A disadvantage is that it requires iteration of each equilibrium
equation as well as iteration among all equations. A method that eliminates itera-
tion of some individual equilibriumequations and reduces the number of iterations
of others (but does not eliminate iteration among all equations) is the analytical
equilibriumiteration (AEI) method (Jacobson 1999b). With this method, most indi-
vidual equilibriumequations are solved analytically instead of iteratively. Solutions
from this method are, to within several decimal places, identical to those from
the MFI method but are obtained signicantly faster. Below, the AEI method is
described for three types of equilibrium equations. Solutions for other types of
reactions are given in Jacobson (1999b).
17.12.1 Reactions of the form D A
For equations of this form, the analytical solution is found by solving the equilib-
rium relation,
c
A.c
c
D.c
=
c
A.0
+ Lx
n
c
D.0
Lx
n
= K
r
(17.87)
where the subscript 0 indicates an initial value, the subscript c indicates a converged
value, Lx
n
is the change in mole concentration required to converge the solution,
K
r
is the equilibrium coefcient (mol mol
1
), which contains activity coefcients,
if treated, and concentrations are in units of mole concentration (mol cm
3
-air).
The analytical solution to this equation is
Lx
n
=
c
D.0
K
r
c
A.0
1 + K
r
(17.88)
which is substituted into c
A.c
= c
A.0
+Lx
n
and c
D.c
= c
D.0
Lx
n
to obtain con-
verged concentrations.
579
Example 17.5
Calculate the equilibrium mole concentrations of O
3
(g) and O
3
(aq) in the pres-
ence of a dilute solution of 1 g m
3
of liquid water (representing the liquid
water content of a cloud) if the only reaction considered is O
3
(g) O
3
(aq),
the atmosphere initially contains 40 g m
3
O
3
(g), and no O
3
(aq), and T =
298.15 K.
SOLUTION
From Appendix Table B.7, the equilibrium coefcient for the reaction
O
3
(g)

O
3
(aq) is K
eq
= 1.13 10
2
mol kg
1
atm
1
. This is converted to
moles of dissolved ozone per mole of gas-phase ozone with
K
r
= K
eq
R
Tm
L
= 0.0113
mol
kg atm
0.08206
Latm
mol K
298.15 K 10
6
kg
L
= 2.76 10
7
mol
mol
where m
L
= 10
6
kg-H
2
O L
1
-air = 1 g m
3
is the liquid water content. The
initial mole concentration of gas-phase ozone is
c
O
3(
g
)
.0
= 40
g
m
3
10
6
m
3
cm
3
10
6
g
g

1
48
mol
g
= 8.33333 10
13
mol
cm
3
The initial aqueous mole concentration is c
O
3(
aq
)
.0
= 0. Substituting these vari-
ables into (17.88) gives
Lx
n
=
8.333333 10
13
mol
cm
3
2.76 10
7
mol
mol
0
1 +2.76 10
7
mol
mol
= 2.299999356 10
19
mol
cm
3
Final concentrations are now
c
O
3
(aq).c
= c
O
3
(aq).0
+Lx
n
= 2.299999356 10
19
mol
cm
3
= 1.104 10
5
g
m
3
c
O
3
(g).c
= c
O
3
(g).0
Lx
n
= 8.33333103 10
13
mol
cm
3
= 39.999988
g
m
3
The solution can be checked by substituting it into (17.87):
c
O
3
(aq).c
c
O
3
(g).c
=
2.299999365 10
19
8.33333103 10
13
= 2.76 10
7
580
17.12 Analytical equilibrium iteration method
Since the ratio yields the equilibrium coefcient, the solution is correct. The
solution implies that relatively little ozone partitions into cloud drops. If
aqueous reactions destroy ozone, though, the removal rate of ozone by clouds
can increase (Chapter 19).
17.12.2 Nonprecipitation reactions of the form D + E A + B
For equations of this form, the analytical solution is found by solving the equilib-
rium relation,
c
A.c
c
B.c
c
D.c
c
E.c
=
(c
A.0
+Lx
n
) (c
B.0
+Lx
n
)
(c
D.0
Lx
n
) (c
E.0
Lx
n
)
= K
r
(17.89)
where K
r
is in units of (mol
2
mol
2
). The analytical solution to this equation is
Lx
n
=
c
A.0
c
B.0
c
D.0
K
r
c
E.0
K
r
+
(c
A.0
+c
B.0
+c
D.0
K
r
+c
E.0
K
r
)
2
4(1 K
r
) (c
A.0
c
B.0
c
D.0
c
E.0
)
2(1 K
r
)
(17.90)
which is substituted into c
A.c
= c
A.0
+Lx
n
. c
B.c
=c
B.0
+Lx
n
. c
D.c
= c
D.0
Lx
n
,
and c
E.c
= c
E.0
Lx
n
to obtain converged concentrations.
17.12.3 Solid reactions of the form D(s) 2A + B
For this type of equation, a solid can form only when
(c
A.0
+2c
D.0
)
2
(c
B.0
+2c
D.0
) > K
r
(17.91)
where K
r
is in units of (mol
3
cm
9
). If this criterion is met, the equilibriumequation
c
2
A.c
c
B.c
= (c
A.0
+2Lx
n
)
2
(c
B.0
+Lx
n
) = K
r
(17.92)
is solved, where c
A.c
= c
A.0
+2Lx
n
. c
B.c
= c
B.0
+Lx
n
. andc
D.c
= c
D.0
Lx
n
,
are the nal, converged concentrations. The term Lx
n
must still be determined. If
the criterion is not met, (17.92) is not solved and the solid concentration is set to
zero.
Due to their complexity, reactions of this form must be solved iteratively. One
solution is with the MFI method. A method that solves twice as fast for this form
of reaction is the NewtonRaphson method. The solution is obtained by rewriting
(17.92) as
f
n
(x) = Lx
3
n.n
+qLx
2
n.n
+rLx
n.n
+s = 0 (17.93)
where q = c
A.0
+c
B.0
. r = c
A.0
c
B.0
+0.25c
2
A.0
. s = c
2
A.0
c
B.0
K
r
, and n is the itera-
tion number. The iterative solution to (17.93) is
Lx
n.n+1
= Lx
n.n
f
n
(x)
f
n
(x)
(17.94)
where f
n
(x) = 3Lx
2
n.n
+2qx
n.n
+r.
581
0
5
10
15
20
25
30
0 5 10 15 20 25 30
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
NO
3
H
2
O(aq) 0.1
SO
4
2
NH
4
+
NaCl concentration ( g m
3
)
Cl
Figure 17.5 Modeled aerosol composition

versus NaCl concentration when the relative
humidity was 90 percent. Other initial con-
ditions were H
2
SO
4
(aq), 10 g m
3
; HCl(g),
0 g m
3
; NH
3
(g), 10 g m
3
; HNO
3
(g),
30 g m
3
; and T = 298 K. NaCl dissolves
and dissociates completely.
0
5
10
15
20
25
0 20 40 60 80 100
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
NH
4
NO
3
(s)
NO
3
Relative humidity (percent)

SO
4
2
NH
4
+

H
2
O(aq) 0.1
(NH
4
)
2
SO
4
(s)
Figure 17.6 Modeled aerosol composition
versus relative humidity. Initial conditions
were H
2
SO
4
(aq), 10 g m
3
; HCl(g), 0
g m
3
; NH
3
(g), 10 g m
3
; HNO
3
(g), 30
g m
3
; and T = 298 K.
17.13 EQUILIBRIUM SOLVER RESULTS
Figures 17.5 and 17.6 show equilibrium aerosol composition determined by a
computer program that used the MFI method. Figure 17.5 illustrates equilibrium
aerosol composition as a function of NaCl concentration. The gure shows that
nearly all ammonia gas dissolved and formed ammonium at low NaCl concen-
tration. As NaCl concentration increased, liquid-water and nitrate concentrations
increased since an increase in NaCl increases hydration, enabling more dissolution
and dissociation of nitric acid. Sulfate existed in the solution at all times.
Figure 17.6 shows the simulated change in aerosol composition as a function of
relative humidity. At lowrelative humidity, aerosol particles contained solid ammo-
nium nitrate and solid ammonium sulfate. When the relative humidity increased
to the MDRH of just less than 62 percent (in this case), which is near the DRH
582
17.14 Nonequilibrium between gases and particles
of ammonium nitrate, liquid water condensed, dissolving both solids. Although its
DRH is about 80 percent, ammonium sulfate dissolved at the MDRH because the
MDRH of a mixture is less than the DRH of any electrolyte in the mixture. As
the relative humidity increased past the MDRH, the liquid-water content of the
solution increased, increasing the rate of dissolution of nitric acid and ammonia
from the gas phase.
17.14 NONEQUILIBRIUM BETWEEN GASES AND PARTICLES
Transfer between a gas and a particle is time-dependent and not instantaneous.
Thus, a real gass saturation vapor pressure over a particles surface does not equal
the gass partial pressure. As such, under many conditions, gasparticle equilibrium
equations should not be used to solve for the partial pressure of a gas. Instead,
transfer between gases and particles should be simulated with time-dependent
(nonequilibrium) growth equations that account for gas dissolution into aerosol
particles.
Generally, the equilibrium assumption between gases and particles holds better
for nucleation- and accumulation-mode particles than for coarse-mode particles
(e.g., Meng and Seinfeld 1996; Capaldo et al. 2000; Pilinis et al. 2000; Fridlind
and Jacobson 2000; Moya et al. 2001). Although the rate of transfer of a gas to the
surface of a large particle is greater than is that to the surface of a small particle,
many more small particles than large particles are generally present. As such, gas
particle uxes are faster to a population of small particles than to a population of
large particles.
In Chapter 16, solutions to nonequilibrium condensational growth and ice
depositional growth equations were given and solved. In the following subsec-
tions, nonequilibrium dissolutional growth equations and a solution to them are
described.
17.14.1 Solution to dissolution growth for nondissociating species
When a gas transfers to a particle surface and dissolves in liquid water on the
surface, the process is called dissolutional growth. Dissolutional growth is treated
here as a nonequilibrium (time-dependent) process. Equations that describe disso-
lutional growth are similar to those that describe condensational growth, (16.67)
and (16.68), except that, for dissolutional growth, the saturation vapor pressure
depends on the molality of gas in solution, whereas for condensational growth, the
saturation vapor pressure is held constant for a given temperature. Thus, conden-
sation can occur on a solid or liquid surface, whereas dissolution must occur in
liquid on a surface.
A saturation vapor pressure was dened earlier as the partial pressure of a
gas over a particle surface when the gas is in equilibrium with the surface at a
given temperature. During dissolution, the saturation vapor pressure of gas A can
be determined from the equilibrium relationship A(g)

A(aq). In terms of an
583
arbitrary nondissociating and nonreacting gas q, the Henrys lawconstant equation
for this reaction is
p
q.s.i
=
m
q.i
H
q
(17.95)
where p
q,s,i
is the saturation vapor pressure of gas q over a particle in size bin
i (atm), m
q,i
is the molality of dissolved gas q in particles of size i (mol kg
1
), and
H
q
= K
eq,q
is the Henrys law constant (mol kg
1
atm
1
). Activity coefcients of
undissociated, dissolved gases are often unity. Equation (17.95) states that, as a gas
dissolves (as its molality increases in solution), the gass saturation vapor pressure
increases.
Saturation vapor pressure (atm) is related to saturation vapor mole concentra-
tion, C
q.s.i
(mol cm
3
), by
p
q.s.i
= C
q.s.i
R
T (17.96)
where R
is the universal gas constant (82.06 cm

3
atmmol
1
K
1
), and Tis absolute
temperature (K). Molality (mol kg
1
) is related to mole concentration in solution
by
m
q.i
=
c
q.i
m
v
c
w.i
(17.97)
where c
q.i
is the mole concentration of dissolved gas q in size bin i (mol cm
3
), c
w,i
is the mole concentration of liquid water in bin i, and m
v
is the molecular weight
of water (0.01802 kg mol
1
). Substituting (17.95) and (17.97) into (17.96) gives
the saturation mole concentration of a nondissociating soluble gas over a dilute
solution as a function of its molality in solution as
C
q.s.i
=
P
q.s.i
R
T
=
m
q.i
R
TH
q
=
c
q.i
m
v
c
w.i
R
TH
q
=
c
q.i
H
q.i
(17.98)
where
H
q.i
= m
v
c
w.i
R
TH
q
(17.99)
is the dimensionless Henrys constant of gas q. If the gas does not dissociate or
chemically react in solution, its time-rate-of-change in concentration due to dis-
solutional growth/evaporation can be found by substituting (17.98) into (16.67).
The result is
dc
q.i.t
dt
= k
q.i.th
C
q.t
S
q.i.th
c
q.i.t
H
q.i.th
(17.100)
where the time subscript for c
q,i,t
is implicit (t) and that for H
q.i .th
is explicit (t h),
allowing for an implicit solution to concentration. The mass transfer rate k
q.i .th
was given in (16.64). The equation for mole conservation between the gas and all
aerosol-particle size bins, corresponding to (17.100), is
dc
q.t
dt
=
N
B
i =1
k
q.i.th
C
q.t
S
q.i.th
c
q.i.t
H
q.i.th
(17.101)
584
Equations (17.100) and (17.101) represent N
B
+ 1 growth/evaporation expres-
sions. Like condensation equations, dissolution equations may be solved with an
iterative ODE solver. Here, a noniterative, mole-conserving, and unconditionally
stable method of solving these equations, called the analytical predictor of disso-
lution (APD) method (Jacobson 1997c), is described.
The noniterative solution to dissolutional growth with the APD scheme is
obtained by assuming that the nal concentration of component q in size bin i
can be calculated by integrating (17.100). The result is
c
q.i.t
=
H
q.i.th
C
q.t
S
q.i.th
+
c
q.i.th
q.i.th
C
q.t
S
q.i.th
exp
hS
q.i.th
k
q.i.th
H
q.i.th
(17.102)
where C
q,t
is currently unknown. As with the APC scheme (Chapter 16), the nal
aerosol and gas concentrations are constrained by
c
q.t
+
N
B
i =1
c
q.i.t
= C
q.th
+
N
B
i =1
c
q.i.th
(17.103)
Substituting (17.102) into (17.103) and solving for C
q,t
give
c
q.t
=
C
q.th
+
N
B
i =1
c
q.i.th
1 exp
hS
q.i.th
k
q.i.th
H
q.i.th
1 +
N
B
i =1
q.i.th
S
i.q.th
1 exp
hS
q.i.th
k
q.i.th
H
q.i.th
(17.104)
which is the nal gas concentration. This concentration is substituted back into
(17.102) to give the nal particle component concentration for a time step. The
APD scheme is unconditionally stable, since all nal concentrations are bounded
between 0 and C
tot
, regardless of the time step and liquid water content, and
it is exactly mole- and mass-conserving. Table 17.5 illustrates the unconditional
stability and mass conservation of the APD scheme. An extension of the APD
scheme to multiple aerosol size distributions is discussed in Jacobson (2002).
17.14.2 Solution to acid/base dissolution growth
When a strong acid, such as nitric acid, hydrochloric acid, or sulfuric acid, dissolves
in solution, it usually dissociates, and the resulting anion may react with a cation,
such as ammonium, sodium, calcium, potassium, or magnesium, in solution, to
form a solid. The simultaneous transfer of acids and bases from the gas phase to
particle surfaces results in instantaneous adjustment to pH, particle composition,
585
Table 17.5 Stability test of the APD scheme
C
q.t
c
q.1.t
c
q.2.t
c
q.3.t
Time step size (s) (g m
3
) (g m
3
) (g m
3
) (g m
3
)
0.1 0.769 3.08 3.08 3.08
10 0.769 3.08 3.08 3.08
60 0.769 3.08 3.08 3.08
600 0.769 3.08 3.08 3.08
7200 0.769 3.08 3.08 3.08
Assume a gas dissolves in three particle size bins, where k
q.i.th
= 0.00333,
0.00833, and 0.0117 s
1
for the respective bins. Also, H
q.i
= 4.0, S
q.i.th
=
1.0, and c
q.i.th
= 0 g m
3
for the bins, and C
q.th
= 10 g m
3
. The table
shows concentrations after 2 h predicted by the APD scheme when ve time
step sizes were used. In all cases, the solution was stable and mass-conserving.
and vapor pressure that renders the resulting set of equations stiff and difcult to
solve at long time step without an oscillatory solution.
In this subsection, a method of solving nonequilibrium dissolutional growth of
acids and bases at high and moderate liquid water contents and long time step is
discussed. In Section 17.14.3, a method of solving such equations at low liquid
water contents is described. In both cases, nonequilibrium growth calculations are
coupled with intra-aerosol equilibriumchemistry calculations. The growth portion
of the algorithm described here is referred to as the predictor of nonequilibrium
growth (PNG) scheme. The equilibrium algorithm is EQUISOLV II. Thus, the cou-
pled scheme is PNG-EQUISOLV II (Jacobson 2005b). Other techniques of solving
nonequilibrium acid/base transfer to aerosol particle surfaces are given in Sun and
Wexler (1998), Capaldo et al. (2000), and Pilinis et al. (2000). In Chapter 19, a
method of solving nonequilibriumdissolutional growth with nonequilibriumaque-
ous chemistry is given.
The growth equation for an acid gas, such as hydrochloric acid (HCl), that
dissolves then dissociates (in this case to the hydrogen ion, H
+
, and the chloride
ion, Cl
) in solution, can be written as

dc
Cl.i.t
dt
= k
HCl.i.th
(C
HCl.t
S
HCl.i.th
C
HCl.s.i.t
) (17.105)
where the subscript t identies values that are currently unknown, the subscript t
h identies values that are known from the beginning of the time step, and
c
Cl.i.t
= c
HCl(aq).i.t
+c
Cl
.i.t
(17.106)
is the mole concentration (mol cm
3
-air) of dissolved, undissociated hydrochloric
acid plus that of the chloride ion in particles of size i at time t. In addition, C
HCl.s.i.t
is the saturation mole concentration (SMC) over the surface of particles of size
i and is currently unknown, and all other variables were dened previously. A
similar equation can be written for nitric acid and sulfuric acid. Sulfuric acid,
586
though, has such a low SMC that its SMC is held constant here at time t h, and
its growth is solved with the APC scheme, discussed in Chapter 16. Ammonia is
a strong base, and its growth solution is coupled with the growth solutions of all
dissolving acids (e.g., HCl, HNO
3
) and condensing acids (e.g., H
2
SO
4
), as discussed
shortly.
The saturation mole concentration of HCl in (17.105) can be approximated
from the gasaerosol equilibrium equations and their corresponding equilibrium
coefcient relationships,
HCl

HCl(aq)
m
HCl(aq).i
p
HCl.s.i
= H
HCl
mol
kg atm
(17.107)
HCl(aq)

H
+
+Cl
m
H
+
.i
m
Cl
.i
2
i.H
+
.Cl
m
HCl(aq).i
= K
HCl
mol
kg
(17.108)
where r
i.H
+
.Cl
is a mean mixed solute activity coefcient (Section 17.7).

Combining (17.96), (17.97), and (17.106)(17.108) gives the saturation mole
concentration of hydrochloric acid as
C
HCl.s.i
=
c
Cl.i
K
HCl.i
(17.109)
where
K
HCl.i
= H
HCl
1 +
K
HCl
(m
v
c
w.i
)
2
R
T
c
H
+
.i
2
i.H
+
.Cl
(17.110)
(mol cm
3
) is the adjusted equilibrium coefcient determined at the beginning of
the growth step fromknown parameters, including the hydrogen ion concentration.
Substituting (17.109) into (17.105) and adding time subscripts gives the growth
equation for HCl as
dc
Cl.i.t
dt
= k
HCl.i.th
C
HCl.t
S
HCl.i.th
c
Cl.i.t
K
HCl.i.th
(17.111)
This equation has two unknowns for each size bin: the gas concentration and the
total dissolved chlorine concentration, and it is identical in form to (17.100).
The hydrochloric acid gas concentration is linked to dissolved chlorine concen-
trations among N
B
particle size bins with the mole conservation relationship,
C
HCl.t
+
N
B
i =1
c
Cl.i.t
= C
HCl.th
+
N
B
i =1
c
Cl.i.th
(17.112)
The solution to (17.111) and (17.112) over one time step is similar to that
for a nondissociating species, except that, each time step, the solution among all
growing and dissociating acids is tied to the growth onto aerosol particles of the
base, ammonia, as described shortly.
587
First, integrating (17.111) gives the growth solution for the chloride ion to a
size bin as
c
Cl.i.t
=
K
HCl.i.th
C
HCl.t
S
Cl
.i.th
+
c
Cl.i.th
HCl.i.th
C
HCl.t
S
HCl.i.th
exp
hk
HCl.i.th
S
HCl.i.th
K
HCl.i.th
(17.113)
where the gas concentration in the equation is still unknown. Substituting (17.113)
into (17.112) and solving give the nal gas concentration as
C
HCl.t
=
C
HCl.th
+
N
B
i =1
c
Cl.i.th
1 exp
hk
HCl.i.th
S
HCl.i.th
K
HCl.i.th
1 +
N
B
i =1
HCl.i.th
S
HCl.i.th
1 exp
hk
HCl.i.th
S
HCl.i.th
K
HCl.i.th
(17.114)
Substituting (17.114) back into (17.113) gives the nal total dissolved chlorine
concentration in each size bin.
Growth equations for other semivolatile acid gases (e.g., HNO
3
, CO
2
) are
solved in the same manner as those for HCl. In the case of carbon dioxide, the
adjusted equilibrium coefcient (17.110) must include a second dissociation con-
stant. Growth equations for involatile acid gases (e.g., H
2
SO
4
) are solved for with
the APC scheme (Chapter 16). Following the growth calculation for all acid gases,
the nal gas and aerosol concentrations of ammonia/ammonium for the time step
are solved assuming ammonia equilibrates with all acids simultaneously after the
acids have grown onto each aerosol size bin. This method produces remarkable
stable, nonoscillatory, and accurate solutions at a long time step (e.g., hundreds of
seconds), as shown shortly.
The ammonia/ammoniumsolution is obtained by solving a charge balance equa-
tion within each size bin, a mole balance equation between the gas phase and each
bin, and an equilibriumrelation between the gas phase and each bin, simultaneously
and after the growth step for acids. With this method, the quantity of ammonium
added to each size bin each time step is the exact amount necessary to balance
charge among all ions in solution following acid growth. Because the equilibration
of ammonia is calculated after the diffusion-limited growth of all acids, ammo-
nia growth is effectively a nonequilibrium growth process. The technique allows
smooth solutions and convergence at long time step among all growing species.
The alternative to equilibrating ammonia is to solve ammonia with (17.113) and
588
(17.114), but such a solution, when coupled with the growth solutions for acids, is
oscillatory and inaccurate unless the time step is short (e.g., a few seconds or less).
A solution with a short time step, though, requires substantial computer time in
three dimensions.
With the present method, the charge balance equation for ammonium within
each size bin is
c
NH
+
4
.i.t
+c
H
+
.i.t
+c
.i.t
= 0 (17.115)
where c
NH
+
4
.i.t
and c
H
+
.i.t
, are unknown, and
c
.i.t
= c
NO

3
.i.t
c
Cl
.i.t
c
HSO

4
.i.t
2c
SO
2
4
.i.t
+
q
zc
q.i.th
(17.116)
consists of known values. The terms outside the summation in (17.116) are concen-
trations of the ions, NO
3
, Cl
, HSO
4
, and SO
4
2
, from the end of the current
time step (time t) of acid growth. The dissolutional and condensational growth
solutions gave total nitrate [HNO
3
(aq) + NO
3
], total chloride [HCl(aq) + Cl
],
and total sulfate [S(VI) = H
2
SO
4
(aq) + HSO
4
+ SO
4
2
] in solution. The par-
titioning of total S(VI) into ions following S(VI) growth, for use in (17.116), is
obtained by applying the ratios, HSO
4
/S(VI) and SO
4
2
/S(VI), determined from
the equilibrium calculation at the beginning of the operator-split time interval,
to the total S(VI) following growth. The partitionings of total chloride and total
nitrate to the chloride ion and the nitrate ion are calculated in a similar manner.
The terms inside the summation in (17.116) are mole concentrations (c) multiplied
by charge (z =+/ 1, 2, 3) of all ions, except those outside the summation, that are
present in solution. The exact charge distribution within particles is recalculated
with an internal-aerosol equilibrium calculation, discussed shortly.
The mole balance equation between gas-phase ammonia and the ammonium
ion within each aerosol size bin is
C
NH
3
.t
+
N
B
i =1
(c
NH
3
(aq).i.t
+c
NH
+
4
.i.t
) = C
NH
3
.th
+
N
B
i =1
(c
NH
3
(aq).i.th
+c
NH
+
4
.i.th
)
= C
tot
(17.117)
Finally, the gasaerosol equilibrium relations for ammonia are
NH
3
(g)

NH
3
(aq)
m
NH
3
(aq).i
p
NH
3
= H
NH
3
mol
kg atm
(17.118)
NH
3
(aq) +H
+

NH
+
4
m
NH
+
4
.i
i.NH
+
4
m
NH
3
(aq).i
m
H
+
.i
i.H
+
= K
NH
3
kg
mol
(17.119)
respectively, where the pressure in the denominator of (17.118) is a partial pressure,
rather than a saturation vapor pressure. The single-ion activity-coefcient ratio in
(17.119) is calculated in terms of binary solute activity coefcients in a mixture
589
with, for example,
i.NH
+
4
i.H
+
=
i.NH
+
4

i.NO

3
i.H
+
i.NO

3
=
2
i .NH
+
4
.NO

3
2
i .H
+
.NO

3
=
i.NH
+
4

i.Cl
i.H
+
i.Cl
2
i .NH
+
4
.Cl
2
i .H
+
.Cl
(17.120)
all of which are calculated during the operator-split internal aerosol equilibrium
calculation. In mole concentration units, (17.118) and (17.119) simplify to
c
NH
3
(aq).i
C
NH
3
= H
NH
3
.i
mol
mol
H
NH
3
.i
= H
NH
3
R
Tm
v
c
w.i
(17.121)
c
NH
+
4
.i
c
NH
3
(aq).i
c
H
+
.i
= K
NH
3
.i
cm
3
mol
K
NH
3
.i
= K
NH
3
1
m
v
c
w.i
i.H
+
i.NH
+
4
(17.122)
where all three parameters on the left side of the equation are unknown.
Equations (17.115), (17.117), (17.121), and (17.122) represent a system of 3N
B
+
1 equations and unknowns. These equations are solved exactly as follows.
Combining (17.121) and (17.122) gives an expression for the hydrogen ion
concentration in each size bin as
c
H
+
.i
=
c
NH
+
4
.i
C
NH
3
H
NH
3
.i
K
NH
3
.i
(17.123)
Substituting this expression into (17.115) gives
c
NH
+
4
.i.t
=
c
.i.t
C
NH
3
.t
H
NH
3
.i.th
K
NH
3
.i.th
C
NH
3
.t
H
NH
3
.i.th
K
NH
3
.i.th
+1
(17.124)
where the gas concentration on the right side of this equation is still unknown. This
equation requires c
.i.t
=min(c
.i.t
,0). If c
.i.t
>0, then the size bin, following acid
growth and before re-equilibration, has a net positive charge without H
+
or NH
4
+
,
which is not physically possible unless acid evaporates from the bin in a quantity
greater than the H
+
+ NH
4
+
present in the bin (since at the beginning of each
growth step, the bin is in charge balance). If c
.i.t
= 0 due to such an occurrence,
c
NH
+
4
.i.t
= 0 for the bin and time step.
Substituting (17.124) and c
NH
3
(aq).i.t
= C
NH
3
.t
H
NH
3
.i.th
into the mole balance
equation (17.117), gives
C
NH
3
.t
+
N
B
i =1
C
NH
3
.t
H
NH
3
.i.th
c
.i.t
C
NH
3
.t
H
NH
3
.i.th
K
NH
3
.i.th
C
NH
3
.t
H
NH
3
.i.th
K
NH
3
.i.th
+1
C
tot
= 0
(17.125)
This nonlinear equation has only one unknown, C
NH
3
.t
. It can be solved with a
NewtonRaphson iteration,
C
NH
3
.t.n+1
= C
NH
3
.t.n
f
n
(C
NH
3
.t.n
)
f
n
(C
NH
3
.t.n
)
(17.126)
590
where n is the iteration number and
f
n
(C
NH
3
.t.n
) = C
NH
3
.t.n
+
N
B
i =1
C
NH
3
.t.n
H
NH
3
.i.th
c
.i.t
C
NH
3
.t.n
H
NH
3
.i.th
K
NH
3
.i.th
C
NH
3
.t.n
H
NH
3
.i.th
K
NH
3
.i.th
+1
C
tot
(17.127)
f
n
(C
NH
3
.t.n
) = 1 +
N
B
i =1
NH
3
.i.th
c
.i.t
H
NH
3
.i.th
K
NH
3
.i.th
C
NH
3
.t.n
H
NH
3
.i.th
K
NH
3
.i.th
+1
+
c
.i.t
C
NH
3
.t.n
NH
3
.i.th
K
NH
3
.i.th
C
NH
3
.t.n
H
NH
3
.i.th
K
NH
3
.i.th
+1
(17.128)
This iteration always converges to a positive number and almost always within
12 iterations when the rst guess of the ammonia gas concentration is zero
(C
NH
3
.t.0
= 0). The computer time for iteration is relatively trivial compared with
the computer time saved by taking a long time step for growth among all growing
acids and between growth and equilibriumcalculations. Once C
NH
3
.t
is obtained, it
is substituted back into (17.124) to obtain the updated ammonium concentration,
into c
NH
3
(aq).i.t
= C
NH
3
.t
H
NH
3
.i.th
to update the liquid ammonia concentration, and
into (17.123) with ammonium to obtain the updated pH. However, pH is updated
more completely during the operator-split equilibrium calculation.
Following the ammonia calculation, a new operator-split internal aerosol equi-
librium calculation is performed to recalculate aerosol ion, liquid, and solid com-
position, activity coefcients, and liquid water content, accounting for all species
in solution. After the equilibrium calculation and after particle size and growth
rates have been updated, another time interval of growth is taken.
Figure 17.7 shows results from a simulation in which nonequilibrium growth
was coupled with equilibrium chemistry when the relative humidity was 90 per-
cent. During the simulation period, H
2
SO
4
condensation, HCl and HNO
3
dissolu-
tion, and NH
3
equilibration were solved. Water uptake was calculated during the
operator-split equilibrium calculation. Figures 17.7 (b) and (c) show time-series
of the aerosol concentrations, summed over all size bins, when the operator-split
time interval between growth and equilibrium was 5 s and 300 s, respectively.
The gures also show results from an equilibrium-only calculation. Results from
both operator-split time steps matched each other and the equilibrium solution
fairly accurately, although the longer time step resulted in a small error during
the rst 15 minutes. Equilibrium was reached within about six hours in both
simulations.
17.14.3 Solution to growth in the presence of solid formation
As the liquid water content of an aerosol population decreases, solid formation
increases due to either ion crystallization or gas heterogeneous reaction on par-
ticle surfaces. At high and moderate liquid water contents, solid formation, if it
591
10
1
10
0
10
1
10
2
10
2
10
1
10
0
10
1
10
2
10
3
10
4
0.1 1 10
d
M

(
g

m
3
)

/

d

l
o
g
1
0

D
p
d
N

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0

D
p
p
, m)
Soildust
NaCl(s)
Number conc.
0
5
10
15
20
25
30
0 2 4 6 8 10 12 S
u
m
m
e
d

c
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Time from start (h)
H
2
O 0.1
NH
4
+
NO
3
Cl
S(VI)
Na
+
0
5
10
15
20
25
30
0 2 4 6 8 10 12
S
u
m
m
e
d

c
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3

)
Time from start (h)
H
2
O 0.1
NH
4
+
NO
3
Cl
S(VI)
Na
+
(a)
(b)
(c)
Figure 17.7 (a) Initial model size distribution, consisting of
10 g m
3
NaCl(s) and 20.4 g m
3
of nonreactive soildust,
spread over 60 size bins. (b) Time series of aerosol component
concentrations, summed over all size bins, from a simulation
of dissolutional growth coupled with chemical equilibrium
(solved with PNG-EQUISOLV II) when the operator-split time
interval was h =5 s. (c) Same as (b) but when h =300 s. The cir-
cles are equilibriumsolutions, found fromEQUISOLVII alone.
The initial gas concentrations were 30 g m
3
HNO
3
(g); 10 g
m
3
NH
3
(g); 0g m
3
HCl(g); and 10g m
3
H
2
SO
4
(g). The
relative humidity and temperature were 90 percent and 298 K,
respectively. From Jacobson (2005b).
592
occurs, is represented well by the dissolutional growth/chemical equilibriumscheme
described in Section 17.14.2. In that case, solid formation is calculated during the
operator-split equilibrium calculation between growth time steps. When the total
(over all sizes) aerosol liquid water content decreases to below about 0.01 g m
3
,
though, solid formation by the dissolutional growth/equilibrium procedure pro-
ceeds slowly at long time steps since solids form only during the operator-splitting
equilibrium calculation. If the time interval between growth and equilibrium is
long, the little gas that dissolves in the small amount of water each growth time
step during the interval cannot form much solid when the equilibrium calculation
is nally performed.
The problem of solid formation at low liquid water content is addressed here by
treating HCl and HNO
3
growth as condensation processes when the liquid water
content is low. At low water contents, growth of CO
2
and other acids are still
treated as dissolution processes since (1) such acids are less important with respect
to solid formation than is HCl or HNO
3
, and (2) the dissolution scheme, which
still allows solids to form through the equilibrium calculation, is unconditionally
stable for any nonzero liquid water content.
When acid growth is solved as a condensation process, the saturation mole
concentration (SMC) of a gas is held constant during a time step of growth. In
addition, the condensed gases are treated as liquids, except that their saturation
mole concentrations are determined based on gassolid equilibrium to ensure the
correct amount of acidic gas transfers to particle surfaces for solid formation. The
liquids/ions are assumed to enter into the small (nonzero) amount of liquid water
on the particle surface. Once the acids are grown, the amount of ammonia that
condenses simultaneously is calculated in the same manner as in Section 17.14.2.
Because acids and the base ammonium are now supersaturated in solution with
respect to the solid phase, solids readily formduring the subsequent internal-aerosol
equilibrium calculation.
The SMCs of HCl and HNO
3
are solved here simultaneously by considering the
two gassolid equilibrium reactions
NH
4
NO
3
(s)

NH
4
(g) +HNO
3
(g) (17.129)
NH
4
Cl(s)

NH
4
(g) +HCl(g) (17.130)
Either solid may form within aerosol particles when either (a) the relative humidity
is increasing and less than the DRH of the solid or (b) the relative humidity is
decreasing and less than the CRH of the solid. DRHs and CRHs were given in
Table 17.4. Ammonium nitrate and ammonium chloride also form only when
p
NH
3
p
HNO
3
> K
NH
4
NO
3
(17.131)
p
NH
3
p
HCl
> K
NH
4
Cl
(17.132)
respectively, where the ps are partial pressures of the gases (atm) and the Ks are
gassolid equilibrium coefcients (atm
2
). At equilibrium, the relation between the
SMC and the equilibrium coefcients of the two reactions can be obtained by
substituting (17.96) into (17.131) and (17.132) and rewriting the equation as an
593
equality, giving
C
NH
3
.s.t
C
HNO
3
.s.t
= K
NH
4
NO
3
(R
T)
2
(17.133)
C
NH
3
.s.t
C
HCl.s.t
= K
NH
4
Cl
(R
T)
2
(17.134)
respectively. If (17.131) and (17.132) are satised, SMCs are calculated by solving
(17.133) and (17.134) together with the mole-balance equation
C
NH
3
.th
C
NH
3
.s.t
= C
HNO
3
.th
C
HNO
3
.s.t
+C
HCl.th
C
HCl.s.t
(17.135)
The analytical solution to these three equations and three unknowns for the ammo-
nia gas saturation mole concentration is
C
NH
3
.s.t
=
1
2
C
0
+
1
2
C
2
0
+4[K
NH
4
NO
3
+ K
NH
4
Cl
](R
T)
2
(17.136)
where
C
0
= C
NH
3
.th
C
HNO
3
.th
C
HCl.th
(17.137)
The saturation mole concentrations of HNO
3
and HCl are nowtrivially found from
(17.133) and (17.134), respectively. The solution assumes that all three SMCs are
independent of particle size.
If only ammonium nitrate can form, (17.136) is solved by removing K
NH
4
Cl
and setting C
0
= C
NH
3
.th
C
HNO
3
.th
. If only ammonium chloride can form, the
equation is solved by removing K
NH
4
NO
3
and setting C
0
= C
NH
3
.th
C
HCl.th
.
Once the saturation mole concentrations for HNO
3
and HCl are calculated,
they are substituted back into growth equations of the form (16.67)/(16.68)
and solved with the APC scheme (16.69)(16.72). Once the acids are grown,
the amount of ammonia that condenses simultaneously is calculated with the
method given in Section 17.14.2. Subsequently, an internal-aerosol equilibrium
calculation is performed during which the ions may crystallize to form solids
through the ionsolid equilibrium equations, NH
4
NO
3
(s)

NH
4
+
+ NO
3
and
NH
4
Cl(s)

NH
4
+
+ Cl
. Alternatively, the ions may hydrate liquid water if con-

ditions are right. Thus, the resulting particle can consist of both aqueous and solid
components.
Figure 17.8 shows the time series of aerosol component concentrations, summed
over all size bins, during a 12-hour nonequilibrium growthequilibrium chemistry
calculation when the relative humidity was 10 percent. The initial aerosol dis-
tribution was the same as that in Fig. 17.7(a). During the period, sulfuric acid,
nitric acid, and hydrochloric acid gasaerosol transfer were treated as nonequilib-
riumcondensation processes. Ammonia gas equilibrated with acids after they grew.
Solids formed during the operator-split equilibrium calculation. The gure shows
that the result with a 300 s time step was nearly the same as that with a 5 s step.
17.15 SUMMARY
In this chapter, chemical equilibrium, activity-coefcient, and water content
equations were discussed and applied. Two methods of solving equilibrium,
594
17.15 Summary
1
10
0 2 4 6 8 10 12
S
u
m
m
e
d

c
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3

)
Time from start (h)
NH
4
NO
3
(s)
NaNO
3
(s)
(NH
4
)
2
SO
4
(s)
Na
2
SO
4
(s)
NH
4
Cl(s)
NaCl(s)
1
10
0 2 4 6 8 10 12
S
u
m
m
e
d

c
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3

)
Time from start (h)
NH
4
NO
3
(s)
NaNO
3
(s)
(NH
4
)
2
SO
4
(s)
Na
2
SO
4
(s)
NH
4
Cl(s)
NaCl(s)
(a)
(b)
Figure 17.8 (a) Time series of aerosol component con-
centrations, summed over all size bins, from a simu-
lation of dissolutional growth coupled with chemical
equilibrium (solved with PNG-EQUISOLV II) when
the operator-split time interval between growth and
equilibrium was h = 5 s. Initial gas concentrations
were 30 g m
3
HNO
3
(g); 10 g m
3
NH
3
(g); 0 g
m
3
HCl(g); and 10 g m
3
H
2
SO
4
(g). In addition,
the relative humidity was 10 percent and T = 298 K.
(b) Same as part (a), but with h =300s. FromJacobson
(2005b).
activity-coefcient, and water content equations simultaneously were also given,
and temperature-dependent expressions for mean binary solute activity coef-
cients were shown. Mean mixed solute activity coefcients were calculated from
mean binary solute activity coefcients with a mixing rule. Techniques for solv-
ing nonequilibrium dissolutional growth among multiple size bins and coupling
growth to equilibrium were also discussed. The growth techniques included those
for nondissociating species, dissociating species, and species that form solids. In
cloud drops and aerosol particles containing substantial water, irreversible aqueous
chemical reaction should also be accounted for. Cloud formation and development
are described in Chapter 18. Irreversible chemistry within aerosol particles and
clouds is discussed in Chapter 19.
595
17.16 PROBLEMS
17.1 Write the equilibriumcoefcient expression, including a solute activity coef-
cient term, for the reaction NaCl(s)

Na
+
+ Cl
.
17.2 (a) Calculate the equilibrium coefcient for the reaction NH
4
NO
3
(s)
+ NaCl(s)

NH
4
Cl(s) + NaNO
3
(s) using thermodynamic data from
Appendix Table B.6.
(b) Calculate the equilibrium coefcient for the same reaction by combining
rate coefcient terms fromAppendix Table B.7 for four separate solidion
equilibrium reactions.
17.3 What is the purpose of calculating mean mixed solute activity coefcients?
Why cant binary solute activity coefcients be used for all calculations?
17.4 How are the thermodynamic activities of gases, liquids, ions, and solids
related to their mole concentrations? Write out expressions for each case.
17.5 List the experimental data required to solve an equilibrium problem that
includes gas, solution, and solid phases. Ignore coupling to growth equa-
tions.
17.6 What is the mathematical relationship between Q
d
and Q
n
in Section 17.11
when Lx = 0?
17.7 (a) Solve (17.91) and (17.92) together analytically, excluding the equilib-
rium term, for c
q.i.t
and C
q.t
. Assume only one size bin is present and
all variables, except c
q.i.t
and C
q.t
, are constant. (Hint: C
q.t
+c
q.i.t
=
C
q.th
+c
q.i.th
, since i = 1.)
(b) Discretize the time derivatives in the equations frompart (a) in rst-order,
backward-difference form, and write out the right sides of both equations
implicitly. Find the resulting solutions for c
q.i.t
and C
q.t
.
17.8 Calculate the equilibrium mole concentrations of HNO
3
(g) and HNO
3
(aq)
in the presence of a dilute solution of 2000 g m
3
of liquid water if the
only reaction considered is HNO
3
(g)

HNO
3
(aq), the atmosphere initially
contains 10 g m
3
HNO
3
(g) and no HNO
3
(aq), and T = 298.15 K.
17.9 Calculate the equilibrium mole concentrations of HNO
3
(g) and NO
3
in the
presence of a dilute solution of 3000 g m
3
of liquid water if the only
reaction considered is NH
3
(g) + HNO
3
(g)

NH
+++
NO
3
, the atmosphere
initially contains 10 g m
3
NH
3
(g), 5 g m
3
HNO
3
and no ions, and T =
298.15 K. Use at least seven digits of accuracy.
17.10 Write a computer script to solve the following problem. Assume uniformly
sized atmospheric particles contain dissolved nitric acid (H
+
, NO
3
), dis-
solved hydrochloric acid (H
+
, Cl
), and liquid water. Assume f

r
= 82 per-
cent, T = 298 K, and the mass concentrations of nitrate (NO
3
) and chloride
(Cl
) at equilibrium in each particle are both 4 micrograms per cubic meter

of air.
(a) Convert the units of chloride and nitrate to moles per cubic centimeter
(mole concentration), and determine the mole concentration of H
+
by
charge balance, assuming H
+
is the only cation and Cl
and NO
3
are
596
the only anions in solution. Use the mole balance to determine the mole
concentrations of HNO
3
and HCl dissolved in solution.
(b) Calculate the liquid-water content of air (micrograms of H
2
O per cubic
meter of air) from the water equation using the water activity data in
Table B.10.
(c) Using the liquid-water content, convert all particle component concen-
trations to units of molality (moles per kilogram of water), and calculate
the pH of the mixed solution, assuming molarity equals molality for con-
venience.
(d) Calculate the binary solute activity coefcients,
H
+
.NO

3
and
H
+
.Cl
,
using Pitzers method. For each calculation, assume H
+
NO
3
and H
+
Cl
pairs are alone in solution. Assume m

NO

3
and m
Cl
are the same as
in part (c). Assume m
H
+ = m
NO

3
and m
H
+ = m
Cl
in each of the two
cases, since the pairs are alone in solution.
(e) Assume H
+
NO
3
and H
+
Cl
pairs are alone in solution. Calculate

the gas-phase partial pressures (in atmospheres) of HCl and HNO
3
from
(17.50).
17.11 Assume for some anion B
that HB(g) dissolves in particles of three sizes,

and its effective Henrys constant for each size bin is H
i.HB
= 4. In solution,
HB(aq) dissociates into H
+
and B
with a dissociation constant K

eq
= 8,
written in generic units. Assume that the activity coefcients are unity, and
ignore the Kelvin effect.
(a) Write the Henrys law and dissociation equilibrium-coefcient relation-
ships for each size bin.
(b) Write the charge-balance constraint for each size bin and the mole-
balance equation for B between the gas and aerosol phases. Assume
C
tot
= 10 g m
3
is the total amount of B in the system.
(c) Solve for the equilibrium gas mole concentration of HB(g) and aerosol
mole concentrations of H
+
, B
, and HB(aq) in each size bin.

(d) Describe how you would solve this problem if the gas phase were not
in equilibrium with the particle phase (i.e., write out and describe the
equations you would solve). You do not have to solve the equations.
17.12 Write a computer program to solve the three equilibrium equations
HNO
3
(g)

H
+
+NO

3
NH
3
(g) +HNO
3
(g)

NH
+
4
+NO

3
HCl(g)

H
+
+Cl
plus the water equation with the AEI method. Assume the gases equilibrate
with the ions in a single dilute size bin (all solute activity coefcients are
unity), and T = 298.15 K. Assume initial gas concentrations of nitric acid,
ammonia, and hydrochloric acid are 5, 15, and 3 g m
3
, respectively. Plot
equilibrium concentrations of all species as a function of relative humidity
between 1 and 99 percent.
597
18
Cloud thermodynamics and dynamics
Clouds affect the atmosphere in several ways. They absorb and reect radiation,
modify local air temperatures, pressures, and winds, produce precipitation, mix
and remove gases and particles, and alter photolysis coefcients. In this chapter,
the thermodynamics and microphysics of clouds are discussed. The thermodynamic
energy equation and the vertical momentum equation for a cloud are rst derived.
Cumulus parameterizations for simulating effects of subgrid scale clouds in a model
are then described. Finally, numerical techniques for size-resolved liquid and ice
cloud microphysics are given. The techniques include those for growth of water
vapor onto aerosol particles to form liquid drops and ice crystals, hydrometeor
hydrometeor coagulation, breakup of large liquid drops, contact freezing, hetero-
geneous/homogeneous freezing, evaporation/sublimation, evaporative freezing, ice
crystal melting, and aerosolhydrometeor coagulation. Finally, numerical methods
of treating removal of aerosol particles by rainout and washout and of treating
lightning generation are given.
18.1 FOG AND CLOUD TYPES AND
FORMATION MECHANISMS
Clouds are a type of hydrometeor. A hydrometeor is an ensemble of liquid or solid
water particles suspended in or falling through the air. In this section, cloud types
and their formation mechanisms are described. Additional discussions of clouds
can be found in Cotton and Anthes (1989), Rogers and Yau (1989), and Houze
(1993).
18.1.1 Cloud classication
Clouds form primarily in the troposphere. In the tropics, the highest clouds reach
heights of 18 km. In midlatitudes and at the poles, tropospheric clouds reach
maximum heights of 13 and 8 km, respectively. Water ice and nitric acid ice clouds
can form in the polar stratosphere. These clouds are discussed in Chapter 11. For
purposes of cloud classication, the troposphere is divided into three altitude ranges
or etages, in which clouds of different types (genera) form most frequently. These
etages encompass high, middle, and low altitude ranges, as shown in Table 18.1.
In 1802, Jean Baptiste Lamarck (17441829) proposed a cloud classication
scheme, but the cloud types he suggested were not generally accepted. In 1803,
598
18.1 Fog and cloud types and formation mechanisms
Table 18.1 Altitude range of cloud etages
Range (km)
Etage Polar regions Temperate regions Tropical regions

High 38 513 69
Middle 24 27 28
Low 02 02 02
Source: WMO (1975).
Luke Howard proposed an alternative identication scheme that used Latin roots
in the cloud names. Sheetlike clouds were called stratus (Latin for layer). Puffy
clouds were called cumulus (heap), featherlike clouds were called cirrus (curl of
hair), and rain clouds were called nimbus (violent rain).
Cloud types are now categorized by 10 genera. Table 18.2 describes the genera
and identies the etage(s) in which each genus is most commonly observed. Agiven
cloud belongs to one genus only.
A fog is essentially a cloud touching the ground, but a fog is not classied as a
cloud. Instead, it has its own classication and is dened as a suspension of liquid
water drops that causes visibility to be reduced to less than 1 km (WMO 1975). If
visibility is greater than 1 km, the fog is called a mist. If the fog contains ice crystals
instead of liquid water, it is an ice fog.
Other hydrometeors suspended in the air include rain, supercooled rain, drizzle,
supercooled drizzle, snow, snowgrains, snowpellets, diamond dust, hail, small hail,
ice pellets, drifting snow, blowing snow, and spray. Rain is precipitation of water
drops falling from a cloud. Supercooled rain is liquid rain at a temperature below
0
C. Drizzle is precipitation of water drops with diameter -0.5 mm. Supercooled

drizzle is liquid drizzle at a temperature below 0
C. Snow is precipitation from a

cloud of single or agglomerated ice crystals. Snow grains are opaque ice crystals
of diameter -1 mm. Snow pellets are rounded white and opaque ice particles with
diameters up to 5 mm. Diamond dust is precipitation of small ice crystals from
the clear sky. Hail is precipitation of clumpy, spheroidal ice particles 550 mm in
diameter. Small hail is precipitation of translucent, spherical ice particles near 5 mm
in diameter. Ice pellets are transparent ice particles, either spheroidal or irregular,
with diameter -5 mm.
Hydrometeors afxed to a surface include fog deposits, dew, frozen dew, hoar
frost, rime, glaze, and freezing rain. Fog deposits are fog drops xed to a sur-
face when the surface temperature exceeds 0
C. Dew is liquid water on a surface

produced from condensation of water vapor. Frozen dew forms from dew if the
temperature drops below 0
C. Hoar frost or frost is ice produced by sublimation

of water vapor on a surface. Rime is ice produced by the freezing of supercooled
fog droplets on a surface when the surface temperature is below 0
C. Glaze is
transparent ice that forms when supercooled drizzle or rain freezes on contact with
a surface. Freezing rain occurs when nonsupercooled drizzle or rain freezes on
contact with a surface.
599
Table 18.2 Cloud classication
Genera Description

Etage
Cirrus (Ci) Detached with white, delicate laments, whitish
patches, or narrow bands. Fibrous appearance
and/or silky sheen.
High
Cirrocumulus (Cc) Thin, white patch, sheet, or layer cloud, made of small
elements with grains or ripples, merged or separate,
but regularly arranged.
High
Cirrostratus (Cs) Transparent, whitish cloud veil. Fibrous or smooth,
totally or partly covering the sky. Produces a halo.
High
Altocumulus (Ac) White and/or gray patch, sheet, or layer, generally
with shading. Contains rounded masses, sometimes
brous, which may or may not be merged.
Middle
Altostratus (As) Grayish sheet or layer, brous or uniform, totally or
partly covers the sky. Has parts thin enough to
reveal the Sun vaguely. No halo.
Middle High
Nimbostratus (Ns) Gray, often dark. Contains rain that almost always
reaches the ground. Thick enough to block the Sun.
Low, ragged clouds appear below the cloud.
Low Middle High
Stratocumulus (Sc) Gray and/or whitish, patch, sheet, or layer. Almost
always has dark parts and rolls, which may or may
not be merged.
Low
Stratus (St) Gray cloud layer with uniform base that may produce
drizzle, ice, or snow. When the Sun is visible
through the cloud, its outline is clear. Stratus does
not produce a halo, except at low temperatures.
Low
Cumulus (Cu) Detached clouds, usually dense, with sharp outlines.
Vertically developed as a rising mound, dome, or
tower, of which the upper part often looks like a
cauliower. The sunlit portion appears white. Its
base is dark and horizontal.
Low Middle
Cumulonimbus (Cb) Heavy, dense cloud with great vertical extent, in the
form of a mountain or tower. Part of upper cloud is
smooth, brous, or striated, and almost always
attened, spreading out like an anvil. Under its dark
base, precipitation occurs.
Low Middle High
Source: WMO (1975).
18.1.2 Cloud formation
Clouds form by one of several mechanisms. Clouds that form by surface heating
and free convection are convective clouds. When the ground is exposed to intense
sunlight, air immediately above the ground heats by conduction. The resulting
unstable temperature prole causes buoyancy and lifting (free convection). The
altitude at which lifting by free convection starts is the level of free convection
(LFC). As warm air near the surface rises, it expands and cools dry-adiabatically. If
the buoyant parcel contains water vapor, and if the parcel temperature cools to the
isentropic condensation temperature (ICT), the vapor condenses, forming a cloud
base. The altitude of the cloud base is the lifting condensation level (LCL). Release
600
18.1 Fog and cloud types and formation mechanisms
Cold air
Warm air
Cold air
C
o
l
d

f
r
o
n
t
W
a
rm
a
ir
W
a
rm
fro
n
t
Figure 18.1 Formation of clouds along a cold and a warm
front. The cold air behind a cold front forces warm air
vertically. The warm air behind a warm front slides up
over cold air ahead of the front. In both cases, the rising
air cools and may result in cloud formation.
of latent heat during condensation at the cloud base provides buoyancy for further
lifting, cooling, and cloud development.
Orographic clouds appear when a horizontal wind encounters a topographic
barrier, such as a mountain, which forces the wind to shift slantwise or vertically
(orographic uplifting). As the air rises, it expands and cools adiabatically. If the
parcel rises to the LCL, a cloud forms.
Clouds can also formwhen winds converge horizontally, such as around a center
of a low pressure. Convergence forces air to rise vertically (forced convection).
Again, as the air rises, it expands and cools, eventually forming a cloud.
Clouds also form along weather fronts, as shown in Fig. 18.1. Along a cold
front, cold, dense air pushes warm, moist air vertically, causing the warm air to
expand and cool until condensation occurs. Along a warm front, warm, moist air
slides over cold air and expands and cools until a cloud forms. The types of clouds
that form along a cold front differ from those that form along a warm front.
Ahead of a cold front, cumulonimbus, altocumulus, cirrostratus, and cirrus are
the most prevalent cloud types. In front of a warm front, stratocumulus, stratus,
nimbostratus, altostratus, cirrostratus, and cirrus are the most likely clouds to
form.
Atmospheric stability above the LCL affects cloud type. If a cloud base is below
2 km, and the air is stable above the LCL, the cloud that forms is a cumulus humilis
(cumulus cloud of slight vertical extent), stratus, or stratocumulus. If the atmo-
sphere is unstable up to the tropopause, the cloud that forms is a cumulonimbus.
Some cloud types form from interactions between other cloud types and the envi-
ronment. When the top of a cirrostratus, altostratus, or stratus cloud cools and the
bottomwarms radiatively, the resulting buoyant motions within the cloud produce
pockets of air that rise or sink to form cirrocumulus, altocumulus, or stratocumu-
lus clouds, respectively. Cirrus clouds often form as the top of a cumulonimbus
cloud dissipates.
601
18.1.3 Fog formation
Fogs form in one of several ways. A radiation fog forms when air near the ground
cools radiatively during the night to the dew point. An advection fog, which gener-
ally appears near the coast, forms when the wind advects warm, moist air (usually
ocean air) over a cold surface, cooling the air to the dew point. An upslope fog
forms when warm, moist air ows up a topographical barrier and cools adiabati-
cally to the dew point. The formation of this type of fog is similar to the formation
of an orographic cloud. An evaporation fog forms when warm water evaporates
and the vapor recondenses when it mixes with cool, dry air. Two types of evapora-
tion fogs are steam fogs and frontal fogs. A steam fog occurs when warm surface
water (e.g., from a lake or the ocean) evaporates and the vapor recondenses as it
rises into cooler air, giving the appearance of rising steam. A frontal fog occurs
when warm raindrops evaporate as they fall through a cold air mass and the vapor
recondenses in the cold air. These conditions often exist ahead of an approaching
front.
Fogs generally burn off from the top and bottom simultaneously. Solar energy
evaporates drops at the fog top, and thermal-infrared energy emitted by the ground
evaporates drops at the fog base. When the fog no longer touches the ground, it is
a stratus cloud.
18.2 MOIST ADIABATIC AND PSEUDOADIABATIC
PROCESSES
The extent to which a cloud grows vertically depends on atmospheric stability and
moisture. In Chapter 2, stability in unsaturated air was discussed. Here, the concept
of stability is extended to saturated air.
An unsaturated parcel of air rising dry-adiabatically cools at the rate of
9.8 K km
1
. If condensation occurs as the parcel rises, latent heat is released,
offsetting the dry-adiabatic cooling, typically by +4 K km
1
but by up to +8 K
km
1
in the tropics. When the released latent heat is absorbed by dry air, water
vapor, and liquid water, the process is moist adiabatic. When, by assumption, it is
absorbed by only dry air and water vapor, the process is pseudoadiabatic.
18.2.1 Pseudoadiabatic lapse rate
In this subsection, the pseudoadiabatic lapse-rate equation is derived. During a
pseudoadiabatic ascent, the latent heat per unit mass of air (J kg
1
) released during
condensation is
dQ = L
e
d
v.s
(18.1)
where L
e
is the latent heat of evaporation (J kg
1
) from (2.53), and
v.s
= p
v.s
p
d
is the mass mixing ratio (kilograms of vapor per kilogram of dry air) of water
vapor at saturation over a liquid surface, dened in (2.67). Combining (18.1) with
602
18.2 Moist adiabatic and pseudoadiabatic processes
the rst law of thermodynamics from (2.87) yields
L
e
d
v.s
= c
p.m
dT
a
dp
a
(18.2)
Substituting
a
= 1
a
= R
m
Tp
a
dT =
R
m
T
c
p.m
p
a
dp
a
L
e
c
p.m
d
v.s
(18.3)
Differentiating (18.3) with respect to altitude and combining the result with
p
a
z =
a
g and p
a
=
a
R
m
T yield
T
z
w
=
R
m
T
c
p.m
p
a
p
a
z

L
e
c
p.m
v.s
z
=
g
c
p.m
L
e
c
p.m
v.s
z
(18.4)
where subscript w signies pseudoadiabatic. Equation (18.4) simplies to (2.90),
the equation for the dry adiabatic lapse rate, when
v.s
z = 0. Differ-
entiating
v.s
= p
v.s
p
d
with respect to altitude, then substituting dp
v.s
=
L
e
p
v.s
dTR
v
T
2
(ClausiusClapeyron equation),
v.s
= p
v.s
p
d
, R
= R
v
, and
p
d
z = p
d
gR
T yield
v.s
z
=

p
d
p
v.s
z

p
v.s
p
d
p
d
z
=
L
e
v.s
R
T
2
T
z
+

v.s
g
R
T
(18.5)
Substituting (18.5) and I
d.m
= gc
p.m
the pseudoadiabatic temperature change with altitude as
T
z
w
= I
w
= I
d.m
1 +
L
e
v.s
R
1 +
L
2
e
v.s
R
c
p.m
T
2
(18.6)
where I
w
is the pseudoadiabatic lapse rate (K km
1
).
Example 18.1
Find I
w
if p
d
= 950 hPa and (a) T = 283 K; (b) T = 293 K.
SOLUTION
(a) From (2.62), p
v,s
= 12.27 hPa; from (2.67),
v,s
= 0.00803 kg kg
1
; from
(2.80), c
p.m
= 1011.6 J kg
1
K
1
(assuming the air is saturated with water
vapor); from (2.54), L
e
= 2.4761 10
6
J kg
1
; and from (18.6) I
w
= 5.21 K
km
1
.
(b) WhenT=293 K, p
v,s
=23.37 hPa,
v,s
=0.0153 kg kg
1
, c
p,m
=1017.9 J kg
1
K
1
, L
e
=2.4522 10
6
J kg
1
, andI
w
=4.27 Kkm
1
. Thus, the pseudoadiabatic
lapse rate changes with temperature and pressure.
603
0.8
1
1.2
1.4
1.6
1.8
2
2.2
2 0 2 4 6 8 10 12 14
A
l
t
i
t
u
d
e

(
k
m
)
Temperature (C)

Absolutely
stable
Absolutely
unstable
Conditionally
unstable
1 4
3
2
d,m
w
Figure 18.2 Stability criteria for unsaturated
and saturated air. If air is saturated, the envi-
ronmental temperature prole (thick solid
lines) is compared with the pseudoadiabatic
prole (right dashed line) to determine sta-
bility. Environmental temperature proles 3
and 4 are stable and 1 and 2 are unstable with
respect to saturated air. Proles 2, 3, and 4 are
stable and 1 is unstable with respect to unsat-
urated air. A rising or sinking air parcel fol-
lows the I
d.m
-line when the air is unsaturated
and the I
w
-line when the air is saturated.
18.2.2 Stability criteria
In (2.100), stability criteria were given for unsaturated air. These criteria can be
expanded to account for saturated and unsaturated air with
I
e
> I
d.m
absolutely unstable
I
e
= I
d.m
unsaturated neutral
I
d.m
> I
e
> I
w
conditionally unstable
I
e
= I
w
saturated neutral
I
e
- I
w
absolutely stable
(18.7)
where I
e
= Tz is the environmental lapse rate from Section 2.6.2.1. If, for
example, I
w
= +6.0 K km
1
and I
e
= +8.0 K km
1
(and I
d.m
= 9.8 K km
1
),
the atmosphere is said to be conditionally unstable, meaning it is stable if the air is
unsaturated but unstable if air is saturated (if a cloud can form). If the atmosphere
is absolutely stable, it is stable in both unsaturated and saturated air. Figure 18.2
shows stability criteria for unsaturated and saturated air, and Fig. 18.3 shows
how stability is determined in atmospheric layers when each layer has a different
ambient environmental temperature prole.
Another method of estimating stability is with equivalent potential temperature,
which is the potential temperature a parcel of air would have if all its water vapor
were condensed and if the latent heat released were used to heat the parcel. It can
be found by lifting a parcel from 1000 hPa to low pressure until all water vapor
is condensed out, then lowering the parcel dry-adiabatically back to 1000 hPa, as
shown in Fig. 18.4. In reality, not all water vapor in a parcel condenses because
604
18.2 Moist adiabatic and pseudoadiabatic processes
0
0.5
1
1.5
2
2.5
3
0 5 10 15 20 25
A
l
t
i
t
u
d
e

(
k
m
)
Temperature (C)
1
2
3
4
5
6
e
d,m
w
Figure 18.3 Determination of stability in
multiple layers of air. Layer 1 is absolutely
unstable, layer 2 is absolutely stable and an
inversion, layer 3 is dry neutral, layer 4 is
conditionally unstable, and layers 5 and 6 are
moist neutral. Stability is determined by com-
paring the slope of the environmental temper-
ature prole (solid line) with the slopes of the
dry and pseudoadiabatic lapse-rate proles.
0
0.5
1
1.5
2
2.5
3
3.5
0 5 10 15 20 25 30 35
A
l
t
i
t
u
d
e

(
k
m
)
Temperature (K)
LCL
q
q
w d,m
d
p
p,e
Figure 18.4 Schematic showing the relation-
ship between potential temperature and equiv-
alent potential temperature. Suppose a parcel
with
p
= 298.15 K (25
C) rises to the LCL

dry-adiabatically then to 3 km pseudoadiabat-
ically, at which point it has lost its water vapor.
If the parcel descends dry-adiabatically back to
the surface, its nal potential temperature is
p.e
= 306.15 K (33
C).
the saturation mass mixing ratio of water always exceeds zero. At high altitudes,
though, the additional condensation from an incremental decrease in temperature
is negligible, and the pseudoadiabatic lapse rate approaches the dry adiabatic lapse
rate.
If the air is initially saturated, as it is above the LCL in Fig. 18.4, the equivalent
potential temperature is
p.e

p
exp
L
e
c
p.d
T
v.s
(18.8)
605
where T is the actual initial temperature of the saturated parcel (K), and
v.s
is the
saturation mass mixing ratio of water vapor at that temperature (kg kg
1
). If the
air is initially unsaturated, as it is below the LCL, equivalent potential temperature
is
p.e

p
exp
L
e
c
p.d
T
LCL
(18.9)
where T
LCL
is the temperature (K) of the parcel if it were lifted dry-adiabatically to
the LCL, and
v
is the initial, unsaturated mass mixing ratio of water vapor in the
parcel.
Stability in unsaturated or saturated air can be found from equivalent potential
temperature by dening a variable,

p.e
that equals
p.e
from (18.8) for either satu-
rated or unsaturated air. For unsaturated air, the temperature in (18.8) is the initial
temperature of the environment, not T
LCL
. In saturated air, the temperature is the
saturated-parcel temperature. In both cases,
v.s
is the saturation mixing ratio at
the temperature used. Conceptually,

p.e
is the value of
p.e
in a hypothetically sat-
urated parcel at the temperature of the parcel. The stability criteria in terms of

p.e
are
p.e
z
- 0 saturated unstable
= 0 saturated neutral
> 0 saturated stable
(18.10)
18.3 CLOUD DEVELOPMENT BY FREE CONVECTION
A fog or cloud forms when the air temperature drops below the dew point. Figure
18.5 shows an example of how the environmental, dry-adiabatic, and pseudoadia-
batic lapse rates can be applied together with the dewpoint to estimate the extent of
cloud development. If air at the surface is moist but unsaturated, the environment
is unstable with respect to dry air. If an external forcing, such as forced convection,
occurs, surface air rises. As the parcel rises, the dry-air pressure ( p
d
) in the parcel
decreases because pressure decreases with increasing altitude. If the water-vapor
mixing ratio (
v
) is constant in the parcel, p
v
must decrease with increasing height
to satisfy
v
= p
v
p
d
. Because p
v
decreases, the dew point must also decrease,
since it is an increasing function of p
v
, as seen from (2.68). The decrease in dew
point with increasing altitude is relatively modest (e.g., around 2 K km
1
).
The temperature in an unsaturated rising parcel decreases at the dry adiabatic
lapse rate. When the dewpoint and the parcel temperature meet, the air is saturated,
condensation occurs, and a cloud forms, as shown in Fig. 18.5. The altitude of cloud
formation is the LCL. The temperature at saturation is the isentropic condensation
temperature (ICT) (K), which usually occurs at the LCL. The ICT is more formally
dened as the temperature at which saturation is reached when unsaturated air is
cooled adiabatically at a constant mass mixing ratio of water vapor. The ICTcan be
approximated by substituting T
IC
and p
d.IC
for T
D
and p
d
, respectively, into (2.68),
606
18.3 Cloud development by free convection
0
0.5
1
1.5
2
2.5
3
5 10 15 20 25 30 35
A
l
t
i
t
u
d
e

(
k
m
)
Temperature (C)
LCL
Dew point
of rising bubble
e
w
d,m
w
Figure 18.5 Simplied model of cumulus-cloud for-
mation. An unsaturated surface parcel of air at 32
C
in an unstable environment (with environmental lapse
rate (I
e
) is perturbed and starts rising. As it rises,
it cools at the dry adiabatic lapse rate (I
d,m
). The
dew point in the parcel decreases at a slower rate.
At the lifting condensation level (LCL), the dew point
equals the parcel temperature, and a cloud begins to
form. Because the parcel temperature is still greater
than that of the environment, the parcel continues to
rise but cools at the pseudoadiabatic lapse rate (I
w
).
When the parcel temperature equals that of the envi-
ronment (cloud top), the cloud parcel stops rising.
combining the result with p
d.IC
= p
d.0
(T
IC
T
0
)
1
from (2.96), where p
d.0
is the
surface air pressure (not necessarily 1000 hPa) and T
0
is the surface temperature
(K), then solving for T
IC
. The result is
T
IC

4880.357 29.66 ln
v
p
d.0
T
IC
T
0
1
/
19.48 ln
v
p
d.0
T
IC
T
0
1
/
(18.11)
where
v
is assumed to be constant between the surface and LCL. The solution
must be found iteratively.
Condensation releases latent heat, giving a parcel buoyancy. As a parcel rises
above the LCL, as in Fig. 18.5, its temperature and dew point decrease pseudoadi-
abatically. So long as the temperature of the cloudy air exceeds that of the ambient
air, the parcel remains buoyant and continues to rise and expand. When the cloud
temperature and ambient temperature equalize, the cloud top decelerates. As shown
in Fig. 18.5, the environmental temperature prole, which was unstable below the
cloud base and within part of the cloud, must become stable for the cloud top to
decelerate.
607
18.4 ENTRAINMENT
Entrainment is the mixing of relatively cool, dry air from outside a cloud with
warm, moist air within the cloud, causing a cloud to evaporate and cool at its
edges. Cool air sinks, creating downdrafts at the edges and reducing the height to
which the cloud penetrates. Entrainment may also occur at the top or base of a
cloud. Detrainment is the opposite of entrainment. During detrainment, air leaves
the cloud and mixes with air around the cloud, slightly increasing the moisture and
heat content of the outside air.
A simple model of the effects of entrainment on a cumulus cloud was developed
by Stommel (1947). This model assumes that entrained air continuously enters a ris-
ing cloud fromits sides and instantaneously mixes uniformly throughout the cloud.
In reality, entrainment is not entirely lateral, instantaneous, or continuous (Houze
1993). Stommels model is a useful tool for one-dimensional studies of clouds and
can be used to derive a simplied entrainment term in the thermodynamic energy
equation.
The model assumes that entrainment affects cloud temperatures in two ways.
First, entrainment forces the cloud to expend energy to heat cool, entrained air to
the cloud virtual temperature. The energy (J) used by the cloud for this purpose
is
dQ
1
= c
p.d
(T
v

T
v
)dM
c
(18.12)
where T
v
is the cloud virtual temperature,

T
v
is the ambient virtual temperature,
and dM
c
is the mass of ambient dry air plus water vapor entrained in the cloud.
Second, entrainment forces the cloud to expend energy evaporating liquid water
to maintain saturation of dry, entrained air. This energy loss (J) is
dQ
2
= L
e
(
v.s

v
)dM
c
(18.13)
where
v.s
is the saturation mass mixing ratio of water vapor in the cloud (kg kg
1
),
and
v
is the mass mixing ratio of water vapor outside the cloud.
Third, the entrainment region gains latent heat energy when rising water vapor
condenses. The energy gained (J) is
dQ
3
= M
c
L
e
d
v.s
(18.14)
where M
c
is the total mass of air within an entrainment region of the cloud, which
consists of dry air, water vapor, and liquid water.
The sum of the three sources and sinks of energy is dQ
= dQ
1
+dQ
2
+dQ
3
.
Substituting (18.12)(18.14) into this equation gives the total change of energy in
the entrainment region of a cloud as
dQ
= c
p.d
(T
v

T
v
)dM
c
L
e
(
v.s

v
)dM
c
M
c
L
e
d
v.s
(18.15)
From (2.82) and (2.70), the rst law of thermodynamics for this problem requires
dQ
= M
c
(c
p.d
dT
v
a
dp
a
) (18.16)
608
18.4 Entrainment
Subtracting (18.16) from (18.15) and rearranging give
c
p.d
dT
v
a
dp
a
= [c
p.d
(T
v

T
v
) + L
e
(
v.s

v
)]
dM
c
M
c
L
e
d
v.s
(18.17)
Dividing (18.17) by c
p.d
T
v
and substituting
a
= R
T
v
p
a
result in
dT
v
T
v
c
p.d
dp
a
p
a
=
T
v

T
v
T
v
+
L
e
(
v.s
v
)
c
p.d
T
v
dM
c
M
c
L
e
d
v.s
c
p.d
T
v
(18.18)
Differentiating (18.18) with respect to height and then substituting p
a
z =
a
g
and p
a
=
a
R
T
v
give the virtual temperature change with altitude in an entraining
cloud as
T
v
z
=
g
c
p.d
(T
v

T
v
) +
L
e
c
p.d
(
v.s

v
)
1
M
c
M
c
z

L
e
c
p.d
v.s
z
(18.19)
When no entrainment occurs (dM
c
= 0), (18.19) simplies to (18.4).
Rearranging (2.103) as
T
v
z
=
T
v
v
z
+
R
T
v
c
p.d
p
a
p
a
z
=
T
v
v
z

g
c
p.d
(18.20)
and substituting the result into (18.19) give the change in potential virtual temper-
ature with altitude in an entrainment region of a cloud as
v
z
=
v
T
v
(T
v

T
v
) +
L
e
c
p.d
(
v.s

v
)
1
M
c
M
c
z

v
T
v
L
e
c
p.d
v.s
dz
(18.21)
Multiplying through by dz and dividing through by dt give the time rate of change
of potential virtual temperature in a cloud as
d
v
dt
=
v
T
v
(T
v

T
v
) +
L
e
c
p.d
(
v.s

v
)
E

v
L
e
c
p.d
T
v
d
v.s
dt
(18.22)
where E = (1M
c
) dM
c
dt is the entrainment rate (s
1
) of outside air into the cloud.
If a thermal is modeled as a spherical bubble with radius r
t
0.2z
c
(m), where z
c
is the center altitude (m) of the thermal above its starting point, then the rate of
entrainment can be approximated with (Houze 1993)
E =
1
M
c
dM
c
dt

3
4r
3
t
d
dt
4r
3
t
3
(18.23)
Equation (3.76) gave the thermodynamic energy equation with diabatic source
and sink terms. One such term is energy release due to condensation, dQ
c
/
e
=
L
e
d
v.s
. Two other terms are energy release due to freezing of liquid water and
deposition of vapor to ice. These terms can be quantied as dQ
f
/
m
= L
m
d
L
and
dQ
dp
/
s
= L
s
d
v.I
, respectively, where L
m
is the latent heat of melting from(2.55),
d
L
is the change in mass mixing ratio (kg kg
1
) of liquid water upon freezing, L
s
609
is the latent heat of sublimation from (2.56),
v.I
p
v.I
p
d
is the saturation mass
mixing ratio of water vapor over an ice surface, which is analogous to (2.67), and
d
v.I
is the change in saturation mass mixing ratio upon sublimation.
Adding these and remaining terms from (3.76) to (18.22) gives the thermody-
namic energy equation in a cloud as
d
v
dt
=
v
T
v
(T
v

T
v
) +
L
e
c
p.d
(
v.s

v
)
E+
1
a
(
a
K
h
)
v
+

v
c
p.d
T
v
L
e
d
v.s
dt
L
m
d
L
dt
L
s
d
v.I
dt
+
dQ
solar
dt
+
dQ
ir
dt
(18.24)
Water categories in a cloud include water vapor, liquid water, drizzle, rainwater,
cloud ice, snow, graupel, and hail. Conversion from one form of water to another
results in a gain or loss of energy, accounted for in the latent-heat terms of (18.24).
Bulk parameterizations of the conversion processes are given in Houze (1993),
Fowler et al. (1996), and Pruppacher and Klett (1997). Size-resolved calculations
of these conversion processes are discussed in Section 18.8.
18.5 VERTICAL MOMENTUM EQUATION IN A CLOUD
In a cloud, vertical scalar velocities are affected by local acceleration, gravity, pres-
sure gradients, and turbulence. From (4.75), the vertical momentum equation in
Cartesian-altitude coordinates was
dw
dt
= g
1
a
p
a
z
+
1
a
(
a
K
m
) w (18.25)
If p
a
and
a
are the pressure and density, respectively, of the air outside a cloud, and
if the ambient air is in hydrostatic balance, p
a
z =
a
g. Adding this equation
to (18.25) gives
dw
dt
= g
a

a
a
( p
a
p
a
)
z
+
1
a
(
a
K
m
) w (18.26)
where p
a
p
a
and
a

a
are the deviations of cloud pressure and density from
ambient pressure and density, respectively.
The buoyancy factor is dened as
B =
a

a
a
=
p
a

T
v
p
a
T
v
p
a

T
v
=
T
v
T
v
T
v
+
T
v
T
v
p
a
p
a
p
a

v
(18.27)
(e.g., Rogers and Yau 1989) where T
v
,
v
, and
a
= p
a
R
T
v
are the virtual temper-
ature, potential virtual temperature, and density, respectively, of cloudy air, and
T
v
,

v
, and
a
= p
a
R

T
v
are the analogous variables for ambient air. The approxi-
mation on the right side of (18.27) was obtained by assuming ( p
a
p
a
)
p
a
is small.
If a parcel contains liquid water, as it does above the lifting condensation level,
condensate (condensed water) adds a downward force to the parcel. Condensate
610
18.5 Vertical momentum equation in a cloud
in the surrounding air also adds a downward force to the surrounding air. To allow
for condensate, the buoyancy factor can be modied to
B =
a

a
a
=
v
(1 +
L
)
v
(1 +
L
)
L
(18.28)
where
L
and
L
are the mass mixing ratios of liquid water in the parcel and
ambient air, respectively (kilograms of liquid water per kilogram of dry air). The
second expression assumes the liquid-water content of the ambient air is small
compared with that in a cloud.
Substituting (18.28) into (18.26) gives the vertical momentum equation as
dw
dt
= g
a
( p
a
p
a
)
z
+
1
a
(
a
K
m
)w (18.29)
From (2.40), (4.48), and (5.38),
1
a
p
a
z
= g =
+
z
= c
p.d
v
P
z
(18.30)
Substituting (18.30) into (18.29) for cloudy and ambient air gives the vertical
momentum equation in a cloud as
dw
dt
= g
c
p.d
v
(P

P)
z
+
1
a
(
a
K
m
)w (18.31)
This equation is similar in several respects to (5.4), the nonhydrostatic vertical
momentum equation. As with (5.4), (18.31) can be solved directly, used to derive
a diagnostic equation for nonhydrostatic pressure, or solved implicitly.
If the pressure perturbation and the eddy diffusion term are ignored, (18.31)
becomes
dw
dt
=
dw
dz
dz
dt
=
dw
dz
w = g
= gB (18.32)
where w = dzdt. Rearranging (18.32) gives wdw = gBdz. Integrating this equa-
tion from a reference height z
a
, where the vertical scalar velocity is w
a
, to height z
yields a simplied expression for the vertical scalar velocity in a cloud,
w
2
= w
2
a
+2g
z
z
a
dz = w
2
a
+2g
z
z
a
Bdz (18.33)
In this expression, w is determined from the integral of buoyancy between the base
of the cloud and the altitude of interest.
611
18.6 CONVECTIVE AVAILABLE POTENTIAL ENERGY
Equation (18.33) can be modied to give an expression for the convective available
potential energy (CAPE), which describes the growth potential of a cloud. The
CAPE determines the buoyant stability of the atmosphere and correlates positively
with growth of and rainfall production in cumulus clouds (Zawadski et al. 1981).
CAPE (m
2
s
2
) is dened as
CAPE = g
z
LNB
z
LFC
Bdz g
z
LNB
z
LFC
dz (18.34)
where z
LFC
is the level of free convection (LFC), z
LNB
is the level of neutral buoyancy
(LNB),
v
is the potential virtual temperature of a rising parcel of air, and

v
is the
potential virtual temperature of the environment. The LFCis the altitude at which a
parcel of rising air rst becomes warmer than the environment. It may be below or
above the LCL. The LNB is the altitude near the cloud top at which environmental
and cloud temperatures equalize and the cloud is no longer buoyant.
Example 18.2
Estimate CAPE and w for a 10 km-thick cumulonimbus cloud over the ocean
and over land if
v
v
1.5 K over the ocean,
v
v
8 K over land, and
the average ambient virtual temperature between 0 and 10 km is
v
= 288 K in
both cases.
SOLUTION
From (18.34), CAPE 511 m
2
s
2
for the ocean case and 2725 m
2
s
2
for
the land case. If
L
is ignored, (18.33) gives w = 32 m s
1
for the ocean case
and w = 74 m s
1
for the land case. These values are higher than observed
maximums over the ocean and land, which are 10 and 50 m s
1
, respectively,
in thunderstorm clouds.
18.7 CUMULUS PARAMETERIZATIONS
Clouds form over horizontal scales of tens to hundreds of meters. When a models
horizontal resolution is smaller than this, the vertical momentum equation can be
used to reproduce cloud structure. The vertical momentum equation can also be
used to reproduce much of the structure but not the details of a squall-line convec-
tive systemfor grid resolution of up to 4 km(Weisman et al. 1997). Many mesoscale
and global models have horizontal resolutions of 450 km and 100600 km,
respectively. In both types of models, cloud development is a subgrid-scale phe-
nomenon and must be parameterized.
Several techniques have been developed to estimate the effects of subgrid-
scale cumulus clouds on the model-scale environment. These techniques are
called cumulus parameterizations and require input variables from the model-scale
612
18.7 Cumulus parameterizations
environment. Important model-scale variables used to predict subgrid effects are
horizontal and vertical wind speeds, potential temperatures, and total water mix-
ing ratios. Cumulus parameterizations use these variables to adjust potential tem-
perature, total water, and momentum elds and to predict precipitation rates.
The effects of a cumulus parameterization on the model-scale environment are
feedbacks.
Cumulus parameterizations include moist convective adjustment schemes (e.g.,
Manabe et al. 1965; Miyakoda et al. 1969; Krishnamurti and Moxim 1971;
Kurihara 1973), Kuo schemes (Kuo 1965, 1974; Anthes 1977; Krishnamurti et al.
1980; Molinari 1982), the ArakawaSchubert scheme (Arakawa and Schubert
1974; Lord and Arakawa 1980; Kao and Ogura 1987; Moorthi and Suarez 1992;
Cheng and Arakawa 1997; Ding and Randall 1998), and other schemes (Ooyama
1971; Kreitzberg and Perkey 1976; Fritsch and Chappel 1980; Betts 1986; Betts
and Miller 1986; Frank and Cohen 1987; Tiedtke 1989; Kain and Fritsch 1990;
Emanuel 1991; Grell 1993; Hack 1994; Kain 2004). Cotton and Anthes (1989)
discuss several schemes in detail. Three are briey described below.
Moist convective adjustment schemes are the most basic cumulus parameteri-
zations. In these schemes, the model-scale vertical temperature prole is adjusted
to a critical, stable prole when the relative humidity exceeds a specied value and
the temperature prole is unstable with respect to moist air. During adjustment,
the temperature prole is adjusted to the pseudoadiabatic rate, the large-scale rel-
ative humidity is unchanged, condensed water vapor precipitates, and total moist
enthalpy is conserved.
In Kuo schemes, rainfall fromcumulus convection is assumed to occur following
model-scale convergence of moisture. Part of the moisture condenses, releasing
latent heat and increasing rainfall. The rest is used to increase the relative humidity
of the environment. Cloud dynamics and microphysics are not computed in Kuo
schemes, cloud types are not classied, and the altitudes of cloud bases and tops
cannot be found.
In the ArakawaSchubert scheme the model-scale environment is divided into a
cloud layer, where clouds form, and a subcloud mixed layer. Within the cloud layer,
multiple individual clouds are allowed to form. The sum of the individual clouds
in a column makes up a cloud ensemble. The cloud ensemble occupies a horizontal
area much smaller than the horizontal area of a grid cell. Each cloud in an ensemble
has its own fractional entrainment rate, vertical mass ux across the cloud base,
and cloud top height. The fractional entrainment rate is dened as the entrainment
rate per unit height divided by the vertical mass ux. The ensemble is divided into
subensembles, which consist of clouds with similar fractional entrainment rates.
Equations are derived for each subensemble and summed over all subensembles to
obtain the net effect of the ensemble on the model-scale environment.
Cloud ensembles affect the model-scale environment in two ways. First, when
saturated air containing liquid water detrains (escapes) fromcloud tops and evapo-
rates, it cools the model-scale environment. Evaporation increases the water-vapor
content in the environment. Rates of detrainment differ for different cloud types
and cloud-top heights. Second, cumulus convection, which occurs when clouds
613
grow vertically, induces subsidence between clouds. During subsidence, the model-
scale temperature increases and relative humidity decreases.
An advantage of the ArakawaSchubert scheme is its sophisticated treatment of
subgrid-scale clouds. Disadvantages of the original scheme were the omission of
convective downdrafts and the assumption that all cloud bases existed in the lowest
model layer. Convective downdrafts were added in Cheng and Arakawa (1997).
Some versions of the model nowpermit cloud bases to appear at any altitude (Ding
and Randall 1998).
18.8 CLOUD MICROPHYSICS
Prior sections of this chapter described cloud thermodynamics. Here, cloud micro-
physics, including microphysical interactions of hydrometeor particles with aerosol
particles, is discussed. Numerical techniques are given for condensation/deposition
of water vapor onto aerosol particles to form liquid and ice cloud particles, coagu-
lation among liquid and ice hydrometeors to form precipitation-sized drops, liquid
drop breakup, contact freezing, homogeneous-heterogeneous freezing, evapora-
tion/sublimation of falling drops, evaporative freezing, ice crystal melting, aerosol
hydrometeor coagulation, gas washout, and lightning. Most techniques given here
are described in greater detail in Jacobson (2003).
18.8.1 Condensation and ice deposition onto aerosol particles
Liquid cloud drops rst formwhen water vapor condenses onto pre-existing aerosol
particles in supersaturated air. Liquid drop formation can occur at temperatures
down to 40
C. Liquid drops that exist below 0
C are called supercooled drops.

At subfreezing temperatures, supercooled drop formation competes with ice crys-
tal formation (ice deposition) for the limited amount of water vapor available. In
Chapter 16, the equations for condensation and ice deposition of water vapor onto
a single aerosol particle and a population of particles were derived. Numerical solu-
tions were given for growth onto one aerosol size distribution. Here, the technique
is extended to growth of liquid drops and ice crystals simultaneously onto multiple
aerosol size distributions, such as those described by Table 15.1.
The ordinary differential equations for water vapor condensation/evaporation
and deposition/sublimation onto multiple aerosol size distributions are
dc
L.Ni.t
dt
= k
L.Ni.th
(C
v.t
S
L.Ni.th
C
L.s.th
) (18.35)
dc
I.Ni.t
dt
= k
I.Ni.th
(C
v.t
S
I.Ni.th
C
I.s.th
) (18.36)
respectively. The corresponding vaporhydrometeor mole balance equation is
dC
v.t
dt
=
N
T
N=1
N
B
i =1
k
L.Ni.th
(C
v.t
S
L.Ni.th
C
L.s.th
)
+k
I.Ni.th
(C
v.t
S
I.Ni.th
C
I.s.th
)
(18.37)
where t and t h indicate the end and beginning, respectively, of a time step of
h seconds, subscripts L and I indicate liquid and ice, respectively, subscripts N
614
18.8 Cloud microphysics
and i indicate the aerosol distribution and size bin in the distribution, respectively,
from which the hydrometeor originates, N
T
and N
B
are the number of aerosol size
distributions and the number of size bins in each distribution, respectively, c
L.Ni
and c
I.Ni
are mole concentrations (moles per cubic centimeter of air) of liquid water
and ice, respectively, in size bin i of aerosol distribution N, C
v
is water vapor mole
concentration, C
L.s
and C
I.s
are saturation vapor mole concentrations over at,
dilute liquid water and ice surfaces, respectively (so are independent of the aerosol
distribution and size bin), S
L.Ni
and S
I.Ni
are the saturation ratios at equilibrium
of water vapor over a liquid solution and over an ice surface, respectively, in size
bin i of distribution N, and k
L.Ni
and k
I.Ni
are the growth rates (s
1
) of water
vapor to activated cloud condensation nuclei (CCN) and ice deposition nuclei
(IDN), respectively. CCN were dened in Section 16.2.3.3 and IDN were dened
in Section 16.8. Briey, CCN are aerosol particles that can potentially activate
into cloud drops if the air is supersaturated with respect to liquid over the particle
surface. IDN are aerosol particles that can potentially activate into ice crystals if
the air is supersaturated with respect to ice over the particle surface.
Expressions for the growth rates are
k
L.Ni
=
n
lq.Ni
4r
Ni
D
v
v.L.Ni
F
v.L.Ni
m
v
D
v
v.L.Ni
F
v.L.Ni
L
e
S
L.Ni
C
L.s
h.Ni
F
h.L.Ni
T
L
e
m
v
R
T
1
+1
(18.38)
k
I.Ni
=
n
ic.Ni
4
Ni
D
v
v.I.Ni
F
v.I.Ni
m
v
D
v
v.I.Ni
F
v.I.Ni
L
s
S
I.Ni
C
I.s
h.Ni
F
h.I.Ni
T
L
s
m
v
R
T
1
+1
(18.39)
which were derived from (16.64) and (16.79), respectively. All terms are evaluated
at time t h. In these equations, n
lq.Ni
and n
ic.Ni
are the number concentrations
(particles cm
3
) of CCN and IDN, respectively, in each size bin of each aerosol
distribution.
Other terms in (18.38) and (18.39) include the following: m
v
is the molecular
weight of water vapor (g mol
1
), D
v
is the diffusion coefcient of water vapor in
air (cm
2
s
1
) (from (16.17) with subscript v substituted for subscript q to indicate
water vapor),
Ni
is particle radius (cm),
Ni
is the ice crystal electrical capaci-
tance (cm) (from (16.77)),
v
and
h
are dimensionless factors for water vapor
and energy, respectively, that account for corrections for collision geometry and
sticking probability during growth (from (16.19) and (16.27), respectively), F
v
is
the dimensionless ventilation coefcient for vapor (from (16.24) for liquid and
(16.78) for ice), F
h
is the dimensionless ventilation coefcient for energy (from
(16.31) for liquid and (16.78) for ice), L
e
and L
s
are latent heats of evaporation
and sublimation, respectively (J g
1
),
a
is the thermal conductivity of moist air
(J cm
1
s
1
K
1
), T is temperature (K), and R

(8.31451 J mol
1
K
1
).
The size bins included in (18.36) and (18.37) depend not only on whether CCN
or IDN are present but also on whether conditions for activation of particles in
a given size bin are met. CCN/IDN activation is determined by solving a K ohler
615
equation similar to (16.40) but assuming (a) the Kelvin and solute effects affect the
saturation ratio at equilibrium over liquid water, (b) multiple solutes dissolved in
solution affect the saturation ratio over liquid water, and (c) only the Kelvin effect
affects the saturation ratio at equilibrium over ice. The K ohler equations under
such conditions are
S
L.Ni.th
1 +
2
L.Ni.th
m
v
r
Ni
R
T
L
3m
v
4r
3
Ni
L
n
Ni.th
N
s
q=1
c
q.Ni.th
(18.40)
S
I.Ni.th
1 +
2
I.Ni.th
m
v
r
Ni
R
T
I
(18.41)
where
L.Ni
and
I.Ni
are the surface tensions over liquid and ice, respectively (dyn
cm
1
= g s
2
), m
v
is the molecular weight of water (g mol
1
),
L
and
I
are the
densities of liquid water and ice, respectively (g cm
3
), r
Ni
is particle radius (cm), R
is the universal gas constant (8.314 51 10

7
g cm
2
s
2
mol
1
K
1
), Tis temperature
(K), n
Ni
is the total number concentration of aerosol particles of size i in distribution
N (particles cm
3
-air), N
s
is the number of soluble components in an aerosol
particle, and c
q.Ni
is the mole concentration of soluble component q in aerosol
particles of size i in distribution N. Soluble components include dissociated or
undissociated electrolytes and undissociated soluble molecules. The surface tension
itself depends on solute concentration, as discussed in Section 16.2.3.1.
Equation (18.40) can be rewritten as
S
L.Ni.th
1 +
a
L.Ni.th
r
Ni
b
L.Ni.th
r
3
Ni
(18.42)
where
a
L.Ni.th
=
2
L.Ni.th
m
v
R
T
L
b
L.Ni.th
=
3m
v
4
L
n
Ni.th
N
s
q=1
c
q.Ni.th
(18.43)
Taking the partial derivative of (18.42) with respect to radius and setting it to zero,
solving for the radius, and substituting the result back into (18.42) give the critical
radius for growth (cm) and critical saturation ratio for liquid cloud drop activation
as
r
L.Ni.th
=
3b
L.Ni.th
a
L.Ni.th
S
L.Ni.th
= 1 +
4a
3
L.Ni.th
27b
L.Ni.th
(18.44)
respectively. A CCN activates into a liquid cloud drop under the following
conditions:
CCN activation
r
Ni
> r
L.Ni
and C
v.th
> S
L.Ni.th
C
L.s.th
or
r
Ni
r
L.Ni
and C
v.th
> S
L.Ni.th
C
L.s.th
(18.45)
616
An IDN activates when
IDN activation C
v.th
> S
I.Ni.th
C
I.s.th
(18.46)
Above 0
C, no ice can form and only (18.45) applies. Between 40 and 0
C,
condensation competes with ice deposition, but the saturation vapor mole concen-
tration over ice is less than that over liquid (C
I.s
- C
L.s
), so ice growth is favored.
Below 40
C, no liquid drops can form, and only (18.46) applies.

Example 18.3
Calculate the critical radius and critical supersaturation when a population
of 100 aerosol particles cm
3
, all with initial radius of 0.3 m, contains
2 10
16
mol cm
3
-air of an organic compound that does not dissociate.
Assume T = 298.15 K.
SOLUTION
The initial molality of the organic compound in solution is
m =
c
n
w
=
2 10
16
mol
cm
3
1000
g
kg
100
partic.
cm
3

4
3
3 10
5
cm
3
1
g
cm
3
= 0.0177
mol
kg
From (14.19), the surface tension of liquid water at 298.15 K is 72.225 dyn
cm
1
. From (16.34), the surface tension of water containing 0.0177 mol kg
1
of
an organic compound is 58.31 dyn cm
1
. From (18.43),
a
L
=
2
L
m
v
R
T
L
=
2 58.31
dyn
cm
18.02
g
mol
8.31451 10
7
g cm
2
s
2
mol K
298.15 K 1
g
cm
3
1
g cm
dyns
2
= 8.478 10
8
cm
b
L
=
3m
w
c
4
L
n
=
3 18.02
g
mol
2 10
16
mol
cm
3
4 1
g
cm
3
100
partic.
cm
3
= 8.604 10
18
cm
3
respectively. Finally, from (18.44), the critical radius and critical saturation
ratio are
r
L
=
3b
L
a
L
= 1.744 10
5
cm = 0.174m
S
L
= 1 +
4a
3
L
27b
L
= 1.0032
respectively. Since the radius of the population of aerosol particles exceeds
the critical radius, the population will activate for any saturation ratio greater
than that over an individual particle surface.
617
A noniterative numerical solution to (18.35)(18.37) is obtained by integrating
(18.35) and (18.36) for one size bin over a time step h, yielding
c
L.Ni.t
= c
L.Ni.th
+hk
L.Ni.th
(C
v.t
S
L.Ni.th
C
L.s.th
) (18.47)
c
I.Ni.t
= c
I.Ni.th
+hk
I.Ni.th
(C
v.t
S
I.Ni.th
C
I.s.th
) (18.48)
respectively, where the nal gas mole concentration in both cases, C
v.t
, is currently
unknown. Final hydrometeor and gas concentrations are constrained by the gas
hydrometeor mole-balance equation,
C
v.t
+
N
T
N=1
N
B
i =1
(c
L.Ni.t
+c
I.Ni.t
) = C
v.th
+
N
T
N=1
N
B
i =1
(c
L.Ni.th
+c
I.Ni.th
) = C
tot
(18.49)
Substituting (18.47) and (18.48) into (18.49) and solving for C
v.t
give a generalized
solution for simultaneous condensation/evaporation and deposition/sublimation,
C
v.t
=
C
v.th
+h
N
T
N=1
N
B
i =1
(k
L.Ni.th
S
L.Ni.th
C
L.s.th
+k
I.Ni.th
S
I.Ni.th
C
I.s.th
)
1 +h
N
T
N=1
N
B
i =1
(k
L.Ni.th
+k
I.Ni.th
)
(18.50)
Since C
v.t
from (18.50) can exceed the maximum gas concentration in the system,
C
tot
, C
v.t
must be set to the smaller of itself and C
tot
. Equation (18.50) does not allow
C
v.t
to fall belowzero in any situation. Once C
v.t
is solved, it is substituted back into
(18.47) and (18.48) to give the nal hydrometeor concentration in each size bin of
each distribution. Since (18.47) and (18.48) can result in negative concentrations
or concentrations above the maximum, two limits are placed sequentially after
both equations are solved among all size bins. The rst is
c
L.Ni.t
= MAX(c
L.Ni.t
. 0) c
I.Ni.t
= MAX(c
I.Ni.t
. 0) (18.51)
which is solved for all size bins. The second, shown for liquid (the equation for ice
uses c
I.Ni.t
c
I.Ni.th
instead of c
L.Ni.t
c
L.Ni.th
in the rightmost term) is
c
L.Ni.t
=
C
v.th
C
v.t
+
N
T
N=1
N
B
i =1
MAX[c
L.Ni.th
c
L.Ni.t
. 0]
+MAX[c
I.Ni.th
c
I.Ni.t
. 0]
N
T
N=1
N
B
i =1
MAX[c
L.Ni.t
c
L.Ni.th
. 0]
+MAX[c
I.Ni.t
c
I.Ni.th
. 0]
(c
L.Ni.t
c
L.Ni.th
) (18.52)
where all c
L.N.i.t
and c
I.N.i.t
values on the right side of the equation are determined
from (18.51) after it has been solved for all size bins. The solution in (18.47)
(18.52) is exactly mole conserving between water vapor and all liquid and ice
618
0
200
400
600
800
1000
1200
1400
1600
10 100
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0
D
p
p
, m)
788 hPa
656 hPa
872 hPa
729 hPa
835 hPa
(a)
0
200
400
600
800
1000
1200
1400
1600
10 100
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0
D
p
p
, m)
788 hPa
656 hPa
872 hPa
729 hPa
835 hPa
(b)
Figure 18.6 Condensed water onto 16 CCN dis-
tributions simultaneously in each of ve layers in
which the temperature was above 0
C during a
one-dimensional simulation when cloud activation
was (a) determined from Equation (18.45) and
(b) assumed to occur for all particles >0.2 m
in diameter regardless of their composition. From
Jacobson (2003).
hydrometeors in all size bins of all distributions under all conditions and is nonit-
erative.
Figure 18.6(a) shows a calculation of cloud drop growth at ve altitudes. At each
altitude, water vapor condensed onto the 16 aerosol distributions similar to those
shown in Table 15.1, where CCN activation in each distribution was determined
from (18.45). Liquid drop concentrations were summed over the 16 distributions
in each layer to generate the gure. Figure 18.6(b) shows a similar calculation,
but when all particles >0.2 m in diameter were allowed to activate, regardless of
their compositions. A comparison of the gures suggests that treatment of multiple
distributions, each with different activation characteristics, produced dual peaks
in the resulting cloud drop size distribution. Such dual peaks, which have also been
observed (e.g., Pruppacher and Klett 1997, Figs. 2.25 and 2.12(a)), arose because
different distributions activate at different diameters, causing discontinuities in the
summed size distribution. When the activation diameter was held constant over
all distributions (Fig. 18.6(b)), only one peak arose for each layer. Single peaks are
also widely observed, suggesting that whether one or two peaks exist may depend
619
on the activation properties of the underlying aerosol particles, which depend on
composition, surface tension, and density.
18.8.2 Hydrometeorhydrometeor coagulation
In Chapter 15, the integrodifferential coagulation equation and a numerical
solution to it were given. Here, the solution is extended to liquid and ice
hydrometeor distributions. Hydrometeorhydrometeor coagulation (also called
collision-coalescence) is the main process producing precipitation in warm
clouds.
The scheme described here is semiimplicit, volume conserving, volume-
concentration conserving, positive-denite, unconditionally stable, and noniter-
ative, with no limitation on time step. Other schemes that have been used to solve
coagulation in clouds include those of Tzivion et al. (1987), Hounslowet al. (1988),
Lister et al. (1995), and Bott (2000). All such schemes conserve various properties,
but are explicit; thus, their time step is limited by stability constraints.
The interactions considered here include liquidliquid, iceice, and graupel
graupel self coagulation, liquidice, liquidgraupel, and icegraupel heterocoagu-
lation, and coagulation of aerosol components contained within the hydrometeor
distributions. Whereas the initial sources of liquid and ice hydrometeors in the
atmosphere are condensation and ice deposition, respectively, the initial source of
graupel is iceliquid coagulation. The solution method here accounts for all coagu-
lation interactions simultaneously; thus iceliquid heterocoagulation for example,
is solved together with liquidliquid self-coagulation.
The nal volume concentration of component x, in particles of hydrometeor
distribution Y in bin k at time t due to coagulation is determined with
v
x.Yk.t
=
v
x.Yk.th
+h(T
x.Yk.t.1
+ T
x.Yk.t.2
)
1 +hT
x.Yk.t.3
(18.53)
T
x.Yk.t.1
=
N
H
M=1
P
Y.M
k
j =1
n
Mj.th
k1
i =1
f
Yi.Mj.Yk
Yi.Mj.th
v
x.Yi.t
T
x.Yk.t.2
=
N
H
M=1
N
H
I=1
Q
I.M.Y
k
j =1
n
Mj.th
k
i =1
f
Ii.Mj.Yk
Ii.Mj.th
v
x.Ii.t
T
x.Yk.t.3
=
N
C
j =1
N
H
M=1
[(1 L
Y.M
)(1 f
Yk.Mj.Yk
) + L
Y.M
]
Yk.Mj.th
n
Mj.th
where N
H
is the total number of hydrometeor distributions and N
C
is the number of
size bins in each hydrometeor distribution. In this case, N
H
=3, where distributions
Y = lq = 1 for liquid, Y = ic = 2 for ice, and Y = gr = 3 for graupel. The equa-
tion applies when v
x
is either the total hydrometeor volume concentration (v
T.lq
,
v
T.ic
, or v
T.gr
), the volume concentration of liquid water or ice in a hydrometeor
620
Table 18.3 Values of P
Y.M
and L
Y.M
when
coagulation is treated among liquid (lq), ice (ic),
and graupel (gr) size distributions.
M
Y lq ic gr
P
Y.M
lq 1 0 0
ic 0 1 0
gr 1 1 1
L
Y.M
lq 0 1 1
ic 1 0 1
gr 0 0 0
(v
L.lq
, v
I.ic
, or v
I.gr
), the volume concentration of an aerosol solution incorporated
within a hydrometeor (v
s.lq
, v
s.ic
, or v
s.gr
), or the volume concentration of any indi-
vidual aerosol component incorporated within a hydrometeor (v
q.lq
, v
q.ic
, or v
q.gr
).
If it is total volume concentration, then total hydrometeor number concentration
(n
lq
, n
ic
, or n
gr
) equals total hydrometeor volume concentration divided by the
single-particle volume in the bin. For example,
n
lq.k.t
=
v
T.lq.k.t
lq.k
(18.54)
Alternatively, both sides of (18.53) can be divided by single-particle volume to solve
for number concentration directly, and the result is identical.
In (18.53), f
Ii.Mj.Yk
is the volume fraction of the summed volume of two single
particles, V
Ii.Mj
=
Ii
+
Mj
, from distributions I and M, partitioned to a xed size
bin k of distribution Y. Each V
Ii.Mj
is fractionated between two xed bins in a
volume- and number-conserving manner. The volume fractions are
f
Ii.Mj.Yk
=
Yk+1
V
Ii.Mj
Yk+1
Yk

Nk
V
Ii.Mj
Yk
V
Ii.Mj
-
Yk+1
k - N
C
1 f
Ii.Mj.Yk1

Yk1
- V
Ii.Mj
-
Yk
k > 1
1 V
Ii.Mj

Yk
k = N
C
0 all other cases
(18.55)
Finally, P, Q, and L are either 1 or 0, depending on the coagulation interactions
accounted for. The parameter P
Y.M
= 1 if particles in distribution Y coagulating
with particles in distribution M produce larger particles in distribution Y. The
upper part of Table 18.3 lists the values of P
Y.M
when liquid, ice, and graupel
distributions are considered. The parameter Q
I.M.Y
= 1 if particles in distribution
I coagulating with particles in distribution M produce particles in distribution Y,
and I = M and I = Y. For example, Q
lq,ic,gr
, Q
lq,gr,gr
, Q
ic,lq,gr
, and Q
ic,gr,gr
= 1, but
all other interactions are zero. The parameter L
Y.M
=1 if particles in distribution Y
621
coagulating with particles in distribution Mdo not produce particles in distribution
Y. The lower part of Table 18.3 lists the values of L
Y.M
.
In (18.53), term T
1
accounts for production of larger liquid, ice, and graupel
particles from self-coagulation and production of larger graupel from liquid and
ice heterocoagulation with graupel. Term T
2
accounts for production of graupel
from liquid heterocoagulation with ice. The rst part of term T
3
accounts for self-
coagulation loss of liquid, ice, and graupel to form larger sizes. The second part of
the same termaccounts for loss of liquid and ice by heterocoagulation with graupel
to form more graupel.
Equation (18.53) is solved in a special order. Distributions that have no coag-
ulation production from other distributions (e.g., the liquid and ice distributions)
are solved rst, followed by distributions with production terms from previously
solved distributions (e.g., the graupel distribution). Within each distribution, equa-
tions are solved from bin k = 1 . . . N
C
. The volume concentrations of individual
components within a distribution can be solved in any order. To minimize com-
puter time, all calculations involving a zero value of f, P, Q, or L are eliminated
ahead of time.
The total hydrometeorhydrometeor coagulation kernel (cm
3
particle
1
s
1
) is
Ii.J j
= E
coal.Ii.J j
K
Ii.J j
(18.56)
where E
coal.Ii.J j
is a coalescence efciency (dimensionless) and K
Ii.J j
is a collision
kernel (cm
3
particle
1
s
1
) accounting for several physical processes causing col-
lision in the atmosphere. Collision kernels are given in Section 15.6 for Brownian
motion, convective Brownian motion enhancement, gravitational collection, turbu-
lent inertial motion, turbulent shear, van der Waals/viscous forces, particle shape,
thermophoresis, diffusiophoresis, and electric charge. Coalescence efciencies are
discussed in Section 15.6.8.
18.8.3 Drop breakup
When liquid raindrops grow sufciently large by coagulation, they become unsta-
ble and break up, either spontaneously or upon collision with other hydrometeor
particles. One method of solving for drop breakup is to add breakup terms to
the coagulation solution (e.g., List and Gillespie 1976). This method requires a
fragment probability distribution for the interaction of each hydrometeor pair. A
second method is simply to assume drops break up once they reach a critical size
(e.g., Danielsen et al. 1972). This method requires a breakup distribution, which is
similar to a fragment probability distribution, except that a breakup distribution
does not consider the physics behind breakup. In addition, it assumes all drops
break up once they reach a certain size whereas, in reality, some do not breakup
until they are larger. On the other hand, fragment probabilities contain uncertain-
ties as well.
622
Table 18.4 Polynomial coefcients for drop breakup for use in (18.57)
300 - D 1290 m 1290 - D 5160 m
A
0
0.53098621799986 6.5418838298481
A
1
0.0036655403240035 0.0043878127949574
A
2
0.0000077765141976619 0.0000010066406670884
A
3
2.9695029431377 10
9
7.771123366063 10
11
Source: Jacobson (2003).
0
0.5
1
1.5
2
2.5
0 1000 2000 3000 4000 5000 6000
d
M

/

M
T

d

l
o
g
1
0
D
p
p
, m)
Figure 18.7 Large-drop breakup distribution
from (18.57).
Figure 18.7 shows a raindrop breakup distribution obtained by curve-tting a
measured breakup distribution from Danielsen et al. (1972). The polynomial t
used to produce the gure is
dM
M
T
dlog
10
D
= A
0
+ D[A
1
+ D(A
2
+ DA
3
)] (18.57)
where dM is incremental liquid water mass in a size increment, M
T
is liquid water
mass summed over all sizes, D is particle diameter (m), and Table 18.4 gives the
coefcients used in the equation.
In a model, drop breakup may be assumed to occur when drops exceed 5-mm in
diameter. Upon breakup, mass fractions of the breakup drop are assigned to each
of several hydrometeor size bins. In such a case, the mass fraction of each breakup
drop going to model size bin k of diameter D = d
k
is
f
M.k
=
dM
M
T
dlog
10
d
k
dlog
10
d
k
N
C
k=1
dM
M
T
dlog
10
d
k
dlog
10
d
k
=
[A
0
+d
k
(A
1
+d
k
(A
2
+d
k
A
3
))]d log
10
d
k
N
C
k=1
([A
0
+d
k
(A
1
+d
k
(A
2
+d
k
A
3
))]d log
10
d
k
)
(18.58)
Finally, the number concentration of drops added to each bin of diameter-width
dlog
10
d
k
due to the breakup of a single drop of diameter d
orig
>5mm is Ln
k
=
f
M.k
(d
orig
d
k
)
3
.
623
18.8.4 Contact freezing
Contact freezing is a mechanism by which an aerosol ice contact nucleus (ICN)
collides with the surface of a liquid drop at a subfreezing temperature and causes
the drop to freeze spontaneously. This process can be treated numerically by coag-
ulating size-resolved ICN with size-resolved liquid hydrometeors. As applied here,
the calculation assumes that if an ICN collides with a liquid drop, the drop and its
aerosol inclusions are transferred to the graupel distribution.
The loss of liquid hydrometeor total volume concentration (x = T) and aerosol
component volume concentration (x = q) due to contact freezing is
v
x.lq.k.t
=
v
x.lq.k.th
1 +hT
x.k.t.3
(18.59)
and the corresponding gain of graupel is
v
x.gr.k.t
= v
x.gr.k.th
+v
x.lq.k.t
hT
x.k.t.3
(18.60)
In these equations, the implicit loss coefcient (s
1
) of drops due to their hetero-
coagulation with ICN is
T
x.k.t.3
= F
T
N
C
j =1
N
T
N=1
Yk.Nj.th
F
ICN.Nj
n
Nj.th
(18.61)
where F
ICN.Nj
is the ratio of the number concentration of contact nuclei to that
of total particles in an aerosol size bin, F
T
is a temperature-dependent parameter
that reduces the rate of contact nucleation at high temperatures, and
Yk.Nj.th
is
the kernel (cm
3
particle
1
s
1
) for coagulation between liquid drops and aerosol
particles (Section 15.6). Equations (18.59) and (18.60) are solved in the order
k = 1 . . . N
C
, where N
C
is the number of hydrometeor size bins per hydrometeor
distribution. The nal number concentrations of liquid and graupel hydrometeors
in each bin of each distribution are
n
lq.k.t
=
v
T.lq.k.t
lq.k
(18.62)
n
gr.k.t
=
v
T.gr.k.t
gr.k
(18.63)
respectively. The temperature-dependence parameter in (18.61) is a fraction,
F
T
=
0 T > 3
C
(T +3)15 18 - T - 3
C
1 T - 18
C
(18.64)
where T is in

C. This equation was obtained by noting that Fig. 2 of Pitter
and Pruppacher (1973) shows that kaolinite and montmorillonite contact freeze
100 percent of drops at 18
C and 0 percent of drops at 3
C. The product,
F
T
F
ICN.Nj
n
Nj.th
, is the number concentration of ICN in a given size bin of a given
aerosol size distribution at a given temperature. The fractional number of ICN
(F
ICN.Nj
) should be larger than the fractional number of IDN in each size bin of a
624
distribution since only few aerosol types can serve as IDN but many can serve as
ICN.
Contact freezing freezes a higher proportion of small drops than large drops.
Although the coagulation rate coefcient of an aerosol particle with a liquid drop
generally increases with increasing size, the number concentration of large drops
decreases at a greater rate than the coagulation rate coefcient increases with
increasing size, causing the aerosolliquid drop coagulation rate (rate coefcient
multiplied by number concentrations of colliding particles) to decrease with increas-
ing size.
Even though contact freezing freezes primarily small drops, the number concen-
tration of aerosol particles in the upper troposphere is sufciently small that contact
freezing has relatively little effect on the size distribution of liquid or graupel in the
upper troposphere. In cold, lower-tropospheric regions in which aerosol particle
concentrations are high, though (e.g., in populated northern latitude regions dur-
ing winter), contact freezing is a more important freezing mechanism than in the
upper troposphere.
Example 18.4
Calculate the fraction of the number concentration of 40-m-radius liquid
drops that contact freeze during one hour in the presence of 10
3
aerosol parti-
cles cm
3
. Assume the temperature is 20
C, the coagulation rate coefcient

is 10
4
cm
3
particle
1
s
1
, and a maximumof 5 percent of the aerosol particles
present can serve as contact nuclei.
SOLUTION
The fractional number concentration of liquid drops that freeze is
1
n
lq.t
n
lq.th
= 1
:
T.lq.t
:
T.lq.th
= 1
1
1 +hF
T
F
ICN
= 1
1
1 +3600s 1 0.05 10
4
cm
3
partic. s
10
3
partic.
cm
3
= 0.947
Thus, nearly 95 percent of liquid drops contact freeze within an hour under
the conditions specied.
18.8.5 Homogeneous and heterogeneous freezing
Two other methods of freezing liquid drops are homogeneous and heterogeneous
freezing. Heterogeneous freezing is triggered when an aerosol particle, called an ice
immersion nucleus (IIN) in this case, enters a liquid drop. Once immersed in the
drop, the IINs surface serves as a site for liquid water heterogeneous nucleation to
ice. Once an ice cluster forms on the IINsurface, it grows rapidly, engulng the IIN
and eventually converting all the liquid water in the drop to ice. Often, multiple IIN
625
act simultaneously, enhancing the freezing rate of the drop. Homogeneous freezing
is caused by the homogeneous nucleation of ice crystals within or on the surface
of a liquid drop in the absence of an IIN. Once a homogeneously nucleated cluster
forms, it grows until the entire drop freezes.
One method of modeling homogeneous freezing together with heterogeneous
freezing is to t an equation of drop freezing to laboratory data. This treatment is
uncertain, because it assumes that the composition of a drop in the air is similar
to that in a laboratory experiment, which is often performed with tap water. Thus,
it is uncertain whether the experiments simulated heterogeneous or homogeneous
freezing or both.
Several studies have suggested that the homogeneousheterogeneous freezing
(HHF) temperature is related logarithmically to drop volume (e.g., Bigg 1953; Vali
1971; Pitter and Pruppacher 1973). For heterogeneous freezing, this assumption is
physical since IIN can be randomly distributed within a drop. Under this theory,
the fractional number of drops that freeze at a given temperature and particle size
is
F
Fr.k.t
= min{
lq.k
exp[B(T
c
T
r
)] . 1} (18.65)
where
lq.k
is the volume of a single liquid drop in bin k (cm
3
), T
c
is the temperature
of a population of drops (
C), T
r
is a reference temperature (
C), and B is a tting

coefcient (
C
1
). Since the equation is empirical, units do not equate. Equation
(18.65) suggests that, the lower the temperature and the larger the single-particle
volume, the greater the fractional number of drops that freeze. From the equation,
the equilibrium median freezing temperature (
C), which is the temperature at

which 50 percent of drops freeze (F
Fr
= 0.5), is
T
mf
=T
r
1
B
ln
0.5
lq.k

B=0.475
C
1
; T
r
=0
C T
m
- 15
C
B=1.85
C
1
; T
r
= 11.14
C 15
C T
m
- 10
C
(18.66)
where the parameter values were derived in Jacobson (2003) from a t to data of
Pitter and Pruppacher (1973). Figure 18.8 compares the ts with the original data.
The data suggest that the freezing rate follows logarithmically with drop volume
only for temperatures below 15
C. Above 10
C, no drops freeze. Other values

of B and T
r
derived from laboratory data can be found in Danielsen et al. (1972)
and Orville and Kopp (1977).
Example 18.5
Calculate the fraction of the number concentration of supercooled 10-m- and
100-m-radius liquid drops that freeze due to homogeneous-heterogeneous
freezing at a temperature of 25
C.
626
28
24
20
16
12
10 100 1000 10
4
M
e
d
i
a
n

f
r
e
e
z
i
n
g

t
e
m
p
e
r
a
t
u
r
e

(
C
)
Particle radius (m)
Figure 18.8 Comparison of tted equation
(18.66) for median freezing temperature with
data of Pitter and Pruppacher (1973, Fig. 1),
from which the ts were derived. From Jacob-
son (2003).
SOLUTION
From (18.65),
F
Fr.10m
= 4.19 10
9
cm
3
exp[0.475
C
1
(25
C 0
C)] = 0.000602
F
Fr.100m
= 4.19 10
6
cm
3
exp[0.475
C
1
(25
C 0
C)] = 0.602
Thus, a greater fraction of larger particles than smaller particles freeze at the
same subfreezing temperature.
Equation (18.65) is an equilibrium equation. A time-dependent freezing-rate equa-
tion is
dn
gr.k.t
dt
= n
lq.k.th
lq.k
A exp[B(T
c
T
r
)] (18.67)
where A=10
4
cm
3
s
1
(e.g., Orville and Kopp 1977; Reisin et al. 1996). Integrat-
ing this equation from t = 0 to h gives the number concentration of liquid drops
and graupel particles, respectively, after homogeneousheterogeneous freezing,
as
n
lq.k.t
= n
lq.k.th
(1 F
Fr.k.t
) (18.68)
n
gr.k.t
= n
gr.k.th
+n
lq.k.th
F
Fr.k.t
(18.69)
where
F
Fr.k.t
= 1 exp{hA
lq.k
exp[B(T
c
T
r
)]} (18.70)
is the fractional number of drops that freeze. Component volume concentrations
are similarly calculated for liquid and graupel. Setting F
Fr.k.t
= 0.5 and solving for
627
10
8
10
6
10
4
10
2
10
0
10
2
1 10 100 1000 10
4
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0
D
p
p
, m)
Liquid,
baseline
(with HHF)
Layer below
cloud top
236.988 K
214 hPa
Liquid,
no HHF
Graupel,
no HHF
Graupel,
baseline
(with HHF)
Figure 18.9 Comparison of modeled liquid and
graupel distributions in the layer below cloud
top from a simulation in which homogeneous-
heterogeneous freezing (HHF) was calculated
and one in which it was not. The time over which
freezing occurred was one hour. The temper-
ature is the initial in-cloud temperature. From
Jacobson (2003).
T in (18.70) give the time-dependent median freezing temperature as
T
mf
= T
r
1
B
ln
ln 0.5
hA
lq.k
(18.71)
Equating Equation (18.71) with (18.66) suggests that the time for equilibration
is h
eq
= ln 0.5(0.5A) = 13 862 s when A = 10
4
cm
3
s
1
. This time is much
greater than the time step for typical atmospheric model processes.
Figure 18.9 illustrates the modeled effect on liquid and graupel size distribu-
tions of HHF over one hour. The gure shows that HHF (baseline case) converted
primarily large liquid drops to graupel. The major effect of HHF was to increase
50300 mgraupel particles. Since graupel in this size range merely replaces liquid,
and because much of the graupel sublimates as it falls, the production of graupel
due to HHF had little effect on hydrometeor or aerosol removal below the layer
shown in the gure.
18.8.6 Drop surface temperature and evaporation/sublimation
When a hydrometeor particle falls belowa cloud, where the air is subsaturated with
respect to water vapor, the particle begins to evaporate or sublimate. The shrinkage
rate of a small hydrometeor is sufciently fast that the entire hydrometeor may
dissipate to its aerosol-particle core. Large hydrometeors, though, often survive,
possibly reaching the ground as rain or snow. Evaporation/sublimation reduces the
surface temperature of a hydrometeor particle.
Here, equations are given to describe the nal hydrometeor surface temperature
and evaporation/sublimation rate when a hydrometeor falls through subsaturated
628
air. The procedure is performed by rst calculating the equilibrium surface tem-
perature iteratively then solving evaporation/sublimation using information from
this information. This procedure is based on the method of Beard and Pruppacher
(1971) and Pruppacher and Rasmussen (1979) with modication.
In the case of a liquid drop, the iteration involves solving the following equations:
p
s.n
= p
v.s
(T
s.n
)
Lp
v.n
= 0.3( p
s.n
p
v.n
)
p
f.n
= 0.5( p
s.n
+ p
v.n
)
(18.72)
T
f.n
= 0.5(T
s.n
+ T
a
)
T
s.n+1
= T
s.n
D
v
L
e
a
(1 p
f.n
p
a
)
Lp
v.n
R
v
T
f.n
p
v.n+1
= p
v.n
+Lp
v.n
where the subscript n is the iteration number, T
s.n
is the drop surface tempera-
ture (K) at iteration n, initialized at the ambient temperature T
a
(K), which stays
constant, p
s.n
is the saturation vapor pressure (hPa) over the drop surface, evalu-
ated at the drop surface temperature, Lp
v.n
is an estimated change in water vapor
partial pressure at iteration n, p
f.n
and T
f.n
are average values of water vapor par-
tial pressure and of temperature between the drop surface and ambient air, D
v
is
the diffusion coefcient of water vapor in air (cm
2
s
1
), L
e
is the latent heat of
evaporation (J g
1
),
a
is the thermal conductivity of moist air (J cm
1
s
1
K
1
),
R
v
is the gas constant for water vapor (4614 cm
3
hPa g
1
K
1
), and p
a
is the
ambient air pressure (hPa). The above set of equations is iterated a minimum of
three times. The factor of 0.3 is included in Lp
v.n
to ensure convergence without
overshooting.
The equations differ from Beard and Pruppacher (1971) only in that the present
equations ignore ventilation coefcients for energy and vapor and radiative heating
to eliminate the size-dependence of the iterative calculation, which would add to
computer time without affecting the result signicantly. For example, when size-
dependent parameters are included, the difference in nal drop surface temperature
between a 0.001 and 1000 m drop is only 2.5 percent (0.05 K/2 K) of the mean
drop surface temperature depression when the relative humidity (RH) is 80 percent
and the ambient temperature is 283.15 K. Even at RH=1 percent, the difference is
only 4 percent (0.4 K/10 K). Temperature equations for sublimation are analogous
to those for evaporation.
Figure 18.10 shows the variation in equilibrium drop surface temperature and
other parameters for initial relative humidities of 1 to 100 percent under (a) lower-
(b) mid- and (c) upper-tropospheric conditions. The gure shows that, under lower-
tropospheric conditions (Fig. 18.10(a)), drop surface temperatures can decrease by
as much as 10 K when RH=1 percent. At RH=80 percent, which is more typical
629
270
275
280
285
0
5
10
15
20
0 0.2 0.4 0.6 0.8 1
T
e
m
p
e
r
a
t
u
r
e

(
K
)
V
a
p
o
r

p
r
e
s
.

(
h
P
a
)

a
n
d

f
i
n
a
l

R
H

1
0
Initial relative humidity (fraction)
Initial and final T
a
and initial T
s
Final T
s
Final p
v
= final p
s
Initial p
v
Final RH 10
Initial p
s
(a)
240
241
242
243
244
245
246
247
248
0
0.2
0.4
0.6
0.8
1
0 0.2 0.4 0.6 0.8 1
T
e
m
p
e
r
a
t
u
r
e

(
K
)
V
a
p
o
r

p
r
e
s
.

(
h
P
a
)

a
n
d

f
i
n
a
l

R
H
Initial and final T
a
and initial T
s
Final T
s
Final p
v
= final p
s
Initial p
v
Final RH
Initial p
s
(b)
222
222.5
223
223.5
0
0.02
0.04
0.06
0.08
0.1
0 0.2 0.4 0.6 0.8 1
T
e
m
p
e
r
a
t
u
r
e

(
K
)
V
a
p
o
r

p
r
e
s
.

(
h
P
a
)

a
n
d

f
i
n
a
l

R
H

0
.
0
1
Initial and final T
a
and initial T
s
Final T
s
Final p
v
= final p
s
Initial p
v
Final RH 0.01
Initial p
s
(c)
Figure 18.10 Variation in equilibrium liquid drop surface tem-
perature (T
s
, K), saturation vapor pressure over the drop sur-
face (p
s
, hPa), partial pressure of water away from the drop sur-
face (p
v
, hPa), and relative humidity (RH) near the drop surface
for initial ambient relative humidities of 1 to 100 percent when
(a) the ambient temperature (T
a
) =283.15 Kand the air pressure
(p
a
) =900 hPa (lower-tropospheric conditions), (b) T
a
=245.94
K and p
a
= 440.7 hPa (middle-tropospheric conditions), and
(c) T
a
= 223.25 K and p
a
= 265 hPa (upper-tropospheric con-
ditions). The gure was obtained by solving Equation (18.72).
From Jacobson (2003).
630
10
4
10
3
10
2
10
1
10
0
10
1
10
2
10
3
1 10 100 1000 10
4
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0
D
p
p
, m)
Cloud base
(872 hPa)
Surface,
RH= 75%
below base
Surface,
RH= 99%
below base
Figure 18.11 Effect of evaporation on the size dis-
tribution of precipitation drops reaching the surface
after falling from a cloud base at about 1.25 km
through subsaturated air in which the relative humid-
ity (RH) was 99%and 75%, respectively. FromJacob-
son (2003).
below a cloud, the temperature depression is close to 2 K. Under mid-tropospheric
conditions (Fig. 18.10(b)), supercooled drop temperatures can decrease by as much
as 2.2 K when RH = 1 percent or 1 K when RH = 50 percent. Under upper-
tropospheric conditions (Fig. 18.10(c)), the maximum temperature depression is
about 0.5 K.
From the drop surface temperature and other parameters found iteratively, the
change in drop liquid water volume concentration (cm
3
cm
3
) due to evaporation
is
v
L.lq.k.t.m
= MAX
v
L.lq.k.th
n
lq.k
4r
k
D
v
(1 p
f.nf
p
a
)
( p
v.s.0
p
v.nf
)
L
R
v
T
f.nf
Lz
V
f.lq.k
. 0
m
(18.73)
where the subscripts 0 and nf indicate initial and nal iterated values, respectively,
from (18.72), and the time step over which evaporation occurs is determined as
the layer thickness (cm) divided by the fall speed (cm s
1
) (LzV
f.lq.k
); thus it is the
time that the hydrometeor can last in the layer before falling to the next layer. This
time step is physical since, if the same time step were constant for all drop sizes, big
drops with short lifetimes would evaporate for a longer period than they would
last in the layer. The fall speed of a hydrometeor is discussed in Chapter 20. The
sublimation rates of ice crystals falling through subsaturated air are calculated in
a manner similar to those of liquid drops.
Figure 18.11 shows the modeled effect of evaporation on an idealized size distri-
bution of liquid drops falling about 1.25 km from cloud base to the ground when
the relative humidity is 99 percent and 75 percent. Evaporation eliminated almost
all small drops in both cases and shrank drops more effectively at 75 percent than
at 99 percent relative humidity.
631
10
8
10
6
10
4
10
2
10
0
10 100
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0
D
p
p
, m)
Figure 18.12 Incremental homogeneous-heterogeneous
freezing due to evaporative cooling. The solid line ()
shows a liquid size distribution at an ambient pressure of
214 hPa, an ambient temperature of 236.988 K, and a
relative humidity (RH) of 100%. When the RH decreases to
80%, the drop surface temperature decreases to 236.617 K,
causing 5% more liquid drops to freeze by homogeneous-
heterogeneous nucleation than would otherwise have frozen
at 236.988 K. The incremental 5% is represented by the
dashed line (----).
18.8.7 Evaporative freezing
As liquid drops falling through subsaturated air evaporate, their surfaces cool
(Fig. 18.10). As temperatures decrease below 10
C, the fractional number of

liquid drops that freeze at a given size increases (Equation (18.65)). As such, the
cooling of a liquid drop due to evaporation at its surface is a mechanism of drop
freezing, termed evaporative freezing (Jacobson 2003). The initial temperature of
freezing and the median freezing temperature of small drops are lower than are
those of large drops, but even for small drops, evaporative cooling may enhance
the probability of freezing under the right conditions.
Figure 18.12 illustrates evaporative freezing. The gure shows a liquid hydrom-
eteor size distribution at a relative humidity of 100 percent and T = 236.988 K.
Under these conditions, a portion of the distribution, determined from (18.65),
freezes due to homogeneous-heterogeneous freezing. When the relative humidity
drops to 80 percent, (18.72) predicts a decrease in the surface temperature of
all drops by 0.35 K. At this lower temperature, (18.65) predicts an increase in
the number concentration of drops that freeze by 5 percent. The dashed curve in
Fig. 18.12 shows this incremental change in the number distribution of frozen
drops due to evaporative freezing.
The theory that evaporative cooling at a drops surface as the drop falls through
subsaturated air may enhance its rate of freezing is analogous to the theory, shown
experimentally by Rasmussen and Pruppacher (1982), that evaporative cooling
delays the onset of melting.
632
The evaporative freezing theory also implies that homogeneous-heterogeneous
drop freezing in subsaturated air may occur rst on a drop surface rather than in
the bulk of the drop. Homogeneous-heterogeneous freezing may also occur more
readily at a drop surface than in the bulk of a drop for other reasons. Tabazadeh
et al. (2002), for example, found that ice nucleation is thermodynamically favored
on the surface in comparison with in the bulk of a liquid drop, even when no
temperature gradient exists. Stuart (2002) found that, if a frozen nucleus forms
in the middle of a drop, thin dendrites shoot out to the surface and warm the
entire drop to the melting point within less than 0.1 s. Heat loss by the partially
frozen drop occurs only by conduction and evaporation at the surface, causing the
surface temperature to be cooler than the rest of the drop but warmer than the
drop was originally. Due to the cooler surface relative to the interior, remaining
freezing proceeds from the surface, inward. Thus, although freezing may initiate in
the center, the rate of freezing quickly increases at the surface relative to the interior.
The main difference between the mechanismof Stuart (2002) and pure evaporative
freezing is that, with the former, surface freezing is triggered by freezing within the
drop itself; with the latter, surface freezing is triggered by surface cooling as a drop
falls through subsaturated air. Evaporative freezing can occur on its own or with
one of these other mechanisms to enhance the rate of surface freezing.
18.8.8 Ice crystal melting
If the temperature of an ice crystal or graupel particle increases above the ice
melting point (nominally T
0
= 273.15 K), the frozen hydrometeor begins to melt.
But, when an ice crystal melts in subsaturated air (e.g., when the partial pressure
of water away from the crystal surface is lower than the saturation vapor pressure
over meltwater on the surface), simultaneous evaporation of the meltwater cools
the particle surface, retarding the rate of melting. Thus, the melting temperature
must exceed T
0
for the nal drop surface temperature to equal T
0
. The lower the
relative humidity (partial pressure divided by saturation vapor pressure over liquid
water), the greater the melting temperature.
The melting point of a population of ice particles can be found from
T
melt
= T
0
+MAX
D
v
L
e
a
R
v
p
v.s
(T
0
)
T
0
p
v
T
a
. 0
(18.74)
(Rasmussen and Pruppacher 1982), where p
v.s
(T
0
) is the saturation vapor pressure
over liquid water (hPa) at T
0
, T
a
is the ambient air temperature (K), p
v
is the
ambient partial pressure of water vapor (hPa), D
v
is the diffusion coefcient of
water vapor in air (cm
2
s
1
), L
e
is the latent heat of evaporation (J g
1
),
a
is the
thermal conductivity of moist air (J cm
1
s
1
K
1
), and R
v
is the gas constant for
water vapor (4614 cm
3
hPa g
1
K
1
). When melting occurs, the change in mass (g)
633
of ice in a particle of a given size is
m
ic.Ni.t
= m
ic.Ni.th
MAX
h
4r
Ni
L
m
a
(T
a
T
0
) F
h.I.Ni

D
v
L
e
R
v
p
v.s
(T
0
)
T
0
p
v
T
a
F
v.I.Ni
. 0
(18.75)
(e.g., Rasmussen et al. 1984), where L
m
is the latent heat of melting (J g
1
) and h
is the time step (s). Although the rate of melting increases linearly with increasing
particle radius, mass increases with radius cubed, so small crystals melt completely
much faster than do large crystals.
Example 18.6
Calculate the melting temperature of a frozen particle when the ambient tem-
perature is 273.15 K, total air pressure is 970 hPa, and the relative humidity
is 80 percent.
SOLUTION
From (8.14), the diffusion coefcient of water vapor in air is D
v
= 2.204
10
5
m
2
s
1
. From (2.62), the saturation vapor pressure at 273.15 K is
p
s,v
= 6.112 hPa. From (2.66), the partial pressure of water vapor at a rela-
tive humidity of 80 percent and at 273.15 K is p
v
= 4.89 hPa. From (2.54),
the latent heat of evaporation at 273.15 K is L
e
= 2.501 10
6
J kg
1
. From
(2.5) and (2.6), the thermal conductivities of dry air and water vapor are
d
=
0.023807 and
v
= 0.015606 J m
1
s
1
K
1
, respectively, at 273.15 K. Since
n
v
(n
v
+n
d
) = p
v
p
a
, the thermal conductivity of moist air from (2.7) is then
a
= 0.023746 J m
1
s
1
K
1
. Substituting these values into (18.74) gives the
melting point of an ice crystal as
T
melt
= 273.15 K+
2.204 10
5
m
2
s
2.501 10
6
J
kg
0.023746
J
ms K
461.4
J
kg K
6.112 hPa
273.15 K

4.89 hPa
273.15 K
100 J
hPa m
3
= 275.40 K
Thus, the melting point of an ice crystal at 80 percent relative humidity is
greater than that at 100 percent relative humidity.
18.8.9 Aerosolhydrometeor coagulation: aerosol washout
The two major mechanisms of aerosol removal by precipitation are nucleation
scavenging (rainout) and aerosolhydrometeor coagulation (washout). Rainout
occurs when a CCN activates to form a liquid drop, and the drop coagulates with
other drops to become rain or graupel, which falls to the surface, removing the
634
CCN inclusions. Washout occurs when growing or falling precipitation particles
coagulate with interstitial aerosol particles and fall to the surface, bringing the
aerosol particles with them.
Together, rainout and washout are the most important mechanisms removing
aerosol particles globally. The other mechanisms of aerosol removal, gravitational
settling and dry deposition, are important for very large particles and over long
periods, but not so important relative to rainout and washout for small particles
over short periods.
Aerosolhydrometeor coagulation results in the coalescence of aerosol particles
with liquid or frozen hydrometeors, reducing aerosol particle number. The nal
aerosol-particle total volume concentration (x = T), solution volume concentra-
tion (x = s), or individual component concentration (x = q) in bin k of aerosol
distribution N after one time step of aerosolhydrometeor coagulation can be cal-
culated as
v
x.Nk.t
=
v
x.Nk.th
1 +hT
x.Nk.t.3
T
x.Nk.t.3
=
N
C
j =1
N
H
M=1
Nk.Mj.th
n
Mj.th
(18.76)
where N
H
is the total number of hydrometeor distributions and N
C
is the number
of bins in each hydrometeor distribution. Equation (18.76) is solved for all aerosol
distributions N= 1 . . . N
T
and size bins k = 1 . . . N
B
per distribution. The corres-
ponding nal volume concentrations of hydrometeor particles and their aerosol
inclusions within size bin k of hydrometeor distribution Y after one time step is
v
x.Yk.t
=
v
x.Yk.th
+h(T
x.Yk.t.1
+ T
x.Yk.t.2
)
1 +hT
x.Yk.t.3
T
x.Yk.t.1
=
N
T
N=1
j =1
n
Nj.th
k1
i =1
f
Yi.Nj.Yk
Yi.Nj.th
v
x.Yi.t
T
x.Yk.t.2
=
N
T
N=1
j =1
n
Yj.th
k
i =1
f
Ni.Yj.Yk
Ni.Yj.th
v
x.Ni.t
T
x.Yk.t.3
=
N
B
j =1
N
T
N=1
(1 f
Yk.Nj.Yk
)
Yk.Nj.th
n
Nj.th
(18.77)
which is solved for all hydrometeor distributions Y = 1 . . . N
H
and size bins
k = 1 . . . N
C
. The volume fractions (e.g., f
Yi.Nj.Yk
) in (18.77) are calculated
in the same manner as in (18.55) but, here, represent the fraction of a coagulated
hydrometeoraerosol particle partitioned into a hydrometeor bin. The nal number
concentrations of aerosol particles and hydrometeor particles following aerosol
hydrometeor coagulation are
n
Nk.t
=
v
T.Nk.t
Nk
N= 1. . . N
T
; k = 1. . . N
B
(18.78)
n
Yk.t
=
v
T.Yk.t
Yk
Y = 1. . . N
H
; k = 1. . . N
C
(18.79)
635
0
500
1000
1500
2000
0
5
10
15
20
25
30
35
0.001 0. 01 0. 1 1 10 100
d
n

(
N
o
.

c
m
3
)

/

d

l
o
g
1
0

D
p
d
V

(

m
3

c
m
3
)

/

d

l
o
g
1
0

D
p
p
, m)
Aerosol
volume
Aerosol
number
Figure 18.13 Below-cloud-base aerosol number and
volume concentration (at 902 hpa), summed over 16
size distributions, before (solid lines) and after (short-
dashed lines) aerosolhydrometeor coagulation. The
simulation period was one hour.
respectively. The scheme described above is volume and volume-concentration con-
serving, positive denite, and noniterative. Aerosolhydrometeor coagulation ker-
nels are given in Section 15.6.
Within a cloud, rainout removes >50 percent, whereas washout removes
-0.1 percent of aerosol mass (e.g., Kreidenweis et al. 1997; Jacobson 2003).
Rainout scavenges all large and most midsize particles, which have large mass,
before washout has a chance to remove particles within a cloud. Within a cloud,
rainout may remove 30 to >50% of aerosol number (e.g., Flossmann et al.
1985; Jacobson 2003; Kreidenweis et al. 2003). However, below a cloud, rain-
out does not remove aerosol particles since no activation of new drops occurs
below a cloud. Thus, all below-cloud removal of aerosol particles is by washout.
Figure 18.13 shows the modeled effect of below-cloud washout on the size dis-
tribution of aerosol particles. Washout removed aerosol particles across the entire
size distribution. Because washout removes small particles within a cloud (large
particles in a cloud are removed by rainout) and washout removes small and large
particles below a cloud, washout generally removes more particles within plus
below cloud by number than does rainout (Jacobson 2003).
18.8.10 Gas washout
Precipitation removes soluble gases by washout. As a raindrop falls through air
containing a soluble gas, the gas may dissolve in the drop. As the drop falls further,
more gas will dissolve if the air is supersaturated with the gas (i.e., if the partial
pressure of the gas exceeds its saturation vapor pressure as determined by the
molality of the dissolved gas in rainwater divided by its Henrys law constant). If
the air is undersaturated some of the gas in the drop will evaporate to maintain
saturation at the drop surface. To complicate matters, a gas may dissolve in drops
of one size and evaporate from drops of a different size in a layer.
636
Below, a parameterization for gas washout is given. It accounts for the changing
amount of solute, summed over drops of all sizes, as the drops fall through the air.
The numerical solution is derived by considering the gashydrometeor equilibrium
relation,
c
q.lq.t.m
C
q.t.m
= H
q
R
T
N
C
k=1
p
L.lq.t.k.m
(18.80)
and the gashydrometeor mole-balance equation
C
q.t.m
+c
q.lq.t.m
= C
q.th.m
+c
q.lq.t.m1
Lz
m1
Lz
m
(18.81)
where m is the layer of the model atmosphere (increasing from 1 at the top to
the number of model layers), Lz
m
is the thickness (cm) of a layer, C
q.t.m
is the
mole concentration of gas q (mol cm
3
), c
q.lq.t.m
is the mole concentration of the
dissolved gas, summed over all size bins k = 1 . . . N
C
in the liquid hydrometeor
size distribution, p
L.lq.t.k.m
is liquid precipitation passing through layer m in bin
k (cm
3
cm
3
), R
is the ideal gas constant (0.08206 L atm mol

1
K
1
), H
q
is an
effective Henrys constant for species q (mol L
1
atm
1
), and T is temperature (K).
Equation (18.81) states that the nal gas plus aqueous species concentration after
dissolution/evaporation equals the initial gas concentration in the layer plus the
aqueous concentration in precipitation from the layer above. For the top cloud
layer, c
q.lq.t.m1
= 0. Combining (18.80) and (18.81) gives the nal gas concentra-
tion in any layer m as
C
q.t.m
=
C
q.th.m
+c
q.lq.t.m1
Lz
m1
Lz
m
1 + H
q
R
T
N
C
k=1
p
L.lq.t.m
(18.82)
The nal aqueous mole concentration, used for the calculation in the next layer, is
c
q.lq.t.m
= C
q.th.m
+c
q.lq.t.m1
Lz
m1
Lz
m
C
q.t.m
(18.83)
This solution is exactly mole conserving, noniterative, unconditionally stable, and
positive denite. It states that if rainwater passing through a model layer is already
saturated with dissolved gas (the air is saturated or undersaturated with the gas),
no additional gas can enter the rainwater, but some may evaporate in the current
layer. This mechanism not only removes gases but also transfers gases from a
supersaturated layer to a subsaturated layer. The formulation given here applies
ideally to soluble gases that are not chemically reactive in solution. For reactive
gases, particularly SO
2
, aqueous chemical reactions often occur on time scales
shorter than the time a drop takes to fall from one layer to the next. Irreversible
aqueous chemical reactions occurring within drops are discussed in Chapter 19.
637
18.8.11 Lightning
Lightning is a bolt of electricity that travels between two regions of a cloud, between
two clouds, or between a cloud and the ground. A lightning stroke heats the air to
greater than 30 000
C. The heating causes the air to expand violently, creating a

shock wave that propagates at the speed of sound (about 330 m s
1
), producing
thunder. The energy of the lightning bolt is also intense enough to split molecular
nitrogen (N
2
2
) to N and O, respectively, which together
react to form nitric oxide (NO). Thus, lightning is a natural source of NO in the
atmosphere.
Lightning occurs following the buildup of opposite charges between one part of
a cloud and either another part of the same cloud, a part of another cloud, or the
ground. The opposing charges create an attractive electrostatic force that becomes
strong enough to cause a discharge in electricity.
The force (N) between two point charges Q
0
(C) and Q
1
(C) separated by
distance r
01
(m) is called the electrostatic force and is determined by Coulombs
law,
F
e
=
k
C
Q
0
Q
1
r
2
01
(18.84)
which states that the force between two point charges varies inversely as the square
of the distance separating the charges and is proportional to the magnitude of each
charge. The force is attractive if the point charges have opposite sign. It is repulsive
if they have the same sign. The constant in the equation is Coulombs constant,
k
C
=
1
4
0
= 8.98755 10
9
Nm
2
C
2
= 8.98755 10
11
VcmC
1
(18.85)
which is measured experimentally and is usually written in terms of the permittivity
of free space,
0
= 8.85419 10
12
C
2
N
1
m
2
= 8.85419 10
14
C V
1
cm
1
,
where 1 V = 1 J C
1
and 1 J = 1 N m. The smallest possible charge is that on an
electron, Q = 1.6 10
19
C.
The electric eld strength is dened for a specic location as the sum of elec-
trostatic forces between a point charge at that location and point charges at all
surrounding locations, divided by the charge at the location of interest. For exam-
ple, the electric eld strength at the location of charge Q
0
due to forces between
Q
0
and all charges Q
i
separated by distance r
0i
is
E
f
=
i
F
e.0i
Q
0
=
i
k
C
Q
i
r
2
0i
(18.86)
which has units of N C
1
(V m
1
).
Lightning occurs only when the electric eld strength exceeds the threshold
electric eld strength, E
th
, which ranges from 100 to 400 kV m
1
, with an average
of 300 kV m
1
. In the absence of clouds, the upper atmosphere is charged slightly
638
positive and the surface of the Earth is charged slightly negative, resulting in an
electric eld strength near the surface of only about 130 Vm
1
. The average charge
density at the ground is about 21 electrons cm
3
. The charge density averaged in
the rst kilometer above the ground is about 5 electrons cm
3
(Pruppacher and
Klett 1997).
One theory of how clouds become electried enough for lightning to form is the
theory of particle rebound charging. According to this theory, large liquid drops,
ice crystals, and graupel particles that are polarized with a net negative charge
on their top and positive charge on their bottom may become negatively charged
overall when they collide with and bounce off smaller liquid or ice hydrometeors
that are also polarized. Since large particles fall faster than small particles, a falling
large particle, which has a positive charge, is hit on its bottom by a small particle
in its path. During the brief collision, the positive charge from the bottom of the
large particle transfers to the small particle, giving the large particle a net negative
charge and the small one a net positive charge. Since the large particle is heavy, it
continues falling toward the bottom of the cloud whereas the small particle, which
is light, stays suspended near the middle or top of the cloud.
Rebound charging can be enhanced by the thermo-electric effect. As graupel and
hail fall through the air, they grow by collision and coalescence with supercooled
liquid water drops, which freeze upon contact. The release of latent heat due to the
freezing warms the surface of the graupel or hail. When these warm hydrometeor
particles subsequently collide with and bounce off smaller ice crystals, a net transfer
of positively charged hydrogen ions (H
+
) occurs from the warmer graupel and hail
to the colder crystals. Negatively charged OH
ions try to diffuse in the opposite

direction, but their mobility is lower than that of H
+
ions.
As a thunderstorm cloud gains strength, the charge differential between its top
and base grows, due in part to the thermo-electric rebound charging mechanism,
and intracloud lightning may occur. Since the cloud base is now predominantly
negatively charged, it induces a positive charge on the Earths surface, which is
normally slightly negatively charged. If the charge difference between the cloud
base and ground increases above the threshold electric eld strength, cloud-to-
ground lightning occurs. Several other possible cloud charging mechanisms are
described in Pruppacher and Klett (1997).
Lightning generation by the rebound charging mechanism can be modeled by
considering the collision and bounceoff of size-resolved liquid and solid hydrome-
teors. The rate coefcient for bounceoff (bounceoff kernel) (cm
3
particle
1
s
1
) of
a particle in size bin i of hydrometeor distribution I interacting with a particle in
size bin j of hydrometeor distribution J in any model layer m of a cloud is
B
Ii.J j.m
= (1 E
coal.Ii.J j.m
)K
Ii.J j.m
(18.87)
where E
coal.Ii.J j.m
is the dimensionless coalescence efciency, discussed in Section
15.6.8, and K
Ii.J j.m
is the collision kernel (cm
3
particle
1
s
1
) discussed in Section
15.6. Combining the bounceoff kernel with other terms gives the charge separation
639
rate per unit volume of air (C cm
3
s
1
) in layer m,
dQ
b.m
dt
=
N
H
J =2
N
C
j =1
N
H
I=J
N
C
i =j
B
Ii.J j
(
Ii
n
Ii.t
n
J j.th
+
J j
n
Ii.th
n
J j.t
)
Ii
+
J j
LQ
Ii.J j
m
(18.88)
where N
H
is the number of hydrometeor distributions, n is the number concentra-
tion of hydrometeor particles (particles cm
3
), is the volume of a single particle
(cm
3
particle
1
), and LQ
Ii.J j
is the charge separation per collision (coulombs per
collision). The charge separation per bounceoff due to the thermo-electric rebound
charging mechanism is in the range 15 10
14
C per collision, with an average
of 3.33 10
14
C per collision (Pruppacher and Klett 1997). The charge separated
during a collision is a complex function of the available charge on each colliding
particle, the angle and surface area of collision, the time during which contact
occurs, and the temperature. The charge separation (C per collision), though, must
be limited by
LQ
Ii.J j
= min[3.33 10
16
C. 0.5(Q
Ii
+ Q
J j
)] (18.89)
where
Q
Ii
= 3.333 10
10
2r
2
Ii
(18.90)
(and a similar expression for Q
J j
) is the total charge (C) on a single particle in a
highly electried, precipitating cloud from (15.63), where r
Ii
is the radius in cm.
In the equations discussed in this chapter, N
H
= 3, where distributions 13
are liquid, ice, and graupel distributions, respectively. Since the thermo-electric
rebound charging mechanism occurs among ice crystals and graupel, the rst sum-
mation in (18.88) considers interactions among only distributions 2 and 3. Thus,
(18.88) treats bounceoffs during size-resolved iceice, icegraupel, and graupel
graupel interactions.
Summing the charge separation rate from (18.88) over all cloud layers and
multiplying the result by the cloudy-sky area of the model column give the overall
charge separation rate in the cloudy region of a model column as
dQ
b.c
dt
= F
c
A
cell
K
bot
m=K
top
dQ
b.m
dt
Lz
m
(18.91)
(C s
1
) where F
c
is the cloud fraction in a model column, A
cell
is the total hori-
zontal area of the grid cell (cm
2
), and Lz
m
is the vertical thickness of each model
layer (cm).
The overall charge separation rate is used to calculate the time-rate-of-change
of the in-cloud electric-eld strength (V cm
1
s
1
),
dE
f
dt
=
2k
C
Z
c
Z
2
c
+ R
2
c
dQ
b.c
dt
(18.92)
640
(e.g., Wang and Prinn 2000), where E
f
is in units of V cm
1
, k
C
is Coulombs
constant from (18.85),
Z
c
=
K
bot
m=K
top
Lz
m
(18.93)
is the summed vertical thickness of cloud layers between the bottom layer (K
bot
)
and top layer (K
top
) of the cloud (cm), and
R
c
=
F
c
A
cell
(18.94)
is the horizontal radius of the cloudy region (cm).
The number of intracloud ashes per centimeter per second can nowbe estimated
with
dF
r
dt
=
1
Z
c
E
th
dE
f
dt
(18.95)
where F
r
is the number of ashes per centimeter and E
th
is the threshold electric eld
strength (V cm
1
) dened previously. The cloud-to-ground ash rate is approx-
imately 3045 percent of the intracloud ashrate (Price et al. 1997; Boccippio
et al. 2001).
Example 18.7
Calculate the number of ashes per hour in a cylindrical cloud of radius 0.5
km and thickness 5 km when two populations of particles, with number con-
centrations 1000 and 0.05 particles cm
3
, respectively, are present. Assume
the collision kernel between the populations is 10
4
cm
3
particle
1
s
1
and
that the coalescence efciency is 40 percent.
SOLUTION
The charge separation rate for this problem is
dQ
b.c
dt
= B
1.2
n
1
n
2
LQ
1.2
V
c
= (1 0.4) 10
4
cm
3
partic. s

1000 partic.
cm
3

0.05 partic.
cm
3
3.33 10
14
C
collision
1.25 10
15
cm
3
= 0.125
C
s
where V
c
= 1.25 10
15
cm
3
is the volume of the cylindrical cloud. From
(18.92), the rate of change of the in-cloud electric eld strength is
dE
f
dt
=
2 8.98755 10
11
Vcm
C
5 10
5
cm
(5 10
5
cm)
2
+(0.5 10
5
cm)
2
0.125
C
s
= 0.894
V
cms
641
The number of intracloud ashes per hour in the cloud is then
1
E
th
dE
f
dt
=
1
3000
V
cmash
0.894
V
cms
= 2.98 10
4
ashes
s
= 1.07
ashes
hr
Finally, the number of nitric oxide (NO) molecules produced per cubic centime-
ter of air per second within a lightning region of a cloud or between the cloud and
the ground is
E
NO
=
E
l
F
NO
A
cell
dF
r
dt
(18.96)
where E
l
is the number of joules per lightning ash and F
NO
is the number of NO
molecules produced per joule of energy released. Values of E
l
for cloud-to-ground
lightning range from 1.8 to 11 GJ/ash with a mean of 6.7 GJ/ash (Price et al.
1997). The energy released by intracloud lightning is about 10 percent that of
cloud-to-ground lightning. Values of F
NO
are in the range 515 10
16
molecules
NO/J with an average of 10
17
molecules NO/J (Price et al. 1997). These values
give the number of NO molecules per ash as 9 10
25
1.7 10
27
for cloud-to-
ground lightning and 9 10
24
1.7 10
26
for intracloud lightning. An observed
NO production rate for intracloud lightning is 2.6 10
25
molecules NO/ash
(Skamarock et al. 2003). Estimates of the total NO produced by lightning range
from 5 to 20 Tg-N per year (Price et al. 1997; Bond et al. 2002).
18.9 SUMMARY
In this chapter, cloud, fog, and precipitation classication, formation, develop-
ment, microphysics, and interactions with aerosol particles were discussed. Clouds
form by free convection, orographic uplifting, forced convection, and lifting along
frontal boundaries. Fogs form by radiational cooling, advection of warm moist air
over a cool surface, orographic uplifting, and evaporation of warm water into cold
air. When air rises in a cloud, it expands and cools pseudoadiabatically. Conden-
sation adds energy and buoyancy to a cloud, and entrainment of outside air causes
cooling and downdrafts along its edges. The thermodynamic energy equation for a
cloud takes account of convection, latent-heat release, entrainment, and radiative
effects. Due to strong inertial accelerations in a cloud, the hydrostatic approxima-
tion is not valid; thus, a vertical momentum equation with an inertial acceleration
term is needed. Microphysical processes affecting cloud evolution include con-
densation/deposition, coagulation, drop breakup, contact freezing, homogeneous-
heterogeneous freezing, evaporative freezing, evaporation/sublimation, and melt-
ing. Aerosol particles are removed by clouds and precipitation through rainout
(nucleation scavenging) and washout (aerosolhydrometeor coagulation). Gases
are removed by gas washout. Collision followed by bounceoff among frozen
642
hydrometeors results in charge separation that can produce lightning, a major
natural source of nitric oxide in the atmosphere.
18.10 PROBLEMS
18.1 Calculate the pseudoadiabatic lapse rate when p
d
= 900 hPa and T =
273.15 K.
18.2 Calculate the dew-point lapse rate in an unsaturated parcel of air that rises
from p
d
= 1000 hPa to p
d
= 900 hPa. Assume
v
= 0.003 kg kg
1
and does
not change in the parcel, and the average T
v
= 279 K.
18.3 Calculate the critical radius and critical supersaturation when a population
of 1500 particles cm
3
, all with radius of 0.1 m, contains 1 10
16
mol
cm
3
-air of an organic compound that does not dissociate. Assume T =
298.15 K.
18.4 Calculate the median freezing temperature of a liquid water drop of radius
(a) 10 m and (b) 1000 m. At which size does a higher fraction of drops
freeze at a constant temperature?
18.5 Explain the difference between contact freezing and heterogeneous freezing.
18.6 Calculate the time required for 40 percent of liquid drops to contact freeze
in the presence of 1000 aerosol particles cm
3
if the temperature is 15
C,
the coagulation rate coefcient is 10
4
cm
3
particle
1
s
1
, and a maximum
of 5 percent of the aerosol particles present can serve as contact nuclei.
18.7 Calculate the melting temperature of a frozen particle when the ambient
temperature is 298.15 K, total air pressure is 1013.25 hPa, and the relative
humidity is 50 percent.
18.8 (a) If the number concentrations of particles 10 and 100 m in radius are
1000 and 0.1 particles cm
3
, respectively, the collision kernel between
particles of these two sizes is 10
5
cm
3
particle
1
s
1
, andthe coalescence
efciency is 30 percent, calculate the number of lightning ashes after
one hour in a cloud column 1 km thick and 1 km in radius. Assume
the charge separation per bounceoff is 3.33 10
16
C and the threshold
electric eld strength is 300 kV m
1
.
(b) What is the average number concentration of NO molecules result-
ing from the lightning after one hour assuming no molecules existed
initially?
18.9 Write a computer script to calculate the pseudoadiabatic lapse rate between
z = 0 and 10 km. Assume T = 288 K at the surface and decreases 6.5 K km
1
and the air is saturated with water vapor. Assuming p
a
= 1013.25 hPa at
the surface, use (2.44) to estimate p
a
at each subsequent altitude. Use the
program to estimate the pseudoadiabatic lapse rate at 100-m increments.
18.10 Assume that, at the surface, T=285 K, T
D
=278 K, p
a
=998 hPa, andTz =
11 K km
1
. Write a computer script to estimate p
a
and T at the lifting
643
condensation level. [Hint: First estimate p
v
at the surface from (2.68), which
gives p
d
= p
a
p
v
, then use the result to obtain
v
which stays constant in
the rising parcel of air. Estimate the decrease in p
a
with altitude from (2.42).
Use
v
and p
d
to estimate T
D
at each altitude.]
18.11 Write a computer code to calculate the total number of drops, integrated
over all sizes, that result from the breakup of one 5-mm diameter liquid
water drop assuming the breakup distribution from (18.57). Use a volume
ratio size distribution with 100 size bins between 10 and 5500 m.
18.12 Write a computer program to replicate Fig. 18.10(a).
644
19
Irreversible aqueous chemistry
I
rreversi ble aqueous chemistry, or irreversible chemistry in liquid water, is
an important step in the production of acid rain, acid fog, and acid haze. For
example, it is responsible for converting sulfur dioxide gas to aqueous sulfuric acid
in aerosol particles, cloud drops, and precipitation. Irreversible aqueous reactions
are tightly coupled with reversible (equilibrium) reactions, such as those discussed
in Chapter 17. Since many aqueous reactants originate as gases, aqueous chem-
istry is tightly linked to nonequilibrium gas-to-particle conversion as well. Since
aqueous reactions are described by rst-order, ordinary differential equations, they
are solved with methods similar to those used to solve gas-phase reactions. In this
chapter, the signicance of aqueous chemistry, identication of important aqueous
reactions, and a method of solving aqueous chemistry together with gas-to-particle
conversion are discussed.
19.1 SIGNIFICANCE OF AQUEOUS CHEMICAL REACTIONS
Aqueous chemistry is irreversible chemistry that takes place in liquid water. It
occurs following dissolution of gases into liquid aerosol particles, cloud drops, and
precipitation as follows: First, soluble gases diffuse to and dissolve in liquid water
on a particle surface. Some of the dissolved molecules dissociate reversibly into
ions. The chemicals then diffuse through the water. Irreversible chemical reactions
take place during diffusion. Some of the dissolved molecules and reaction products
may diffuse back to the particle surface and evaporate, whereas others may stay
within the particle.
Gas dissolution and irreversible chemistry affect particle composition at a rate
proportional to the quantity of liquid water present. As such, aqueous chemistry
proceeds faster in cloud drops, which generally have liquid-water contents >10
5
g m
3
, than in aerosol particles, which have liquid-water contents -500 g m
3
.
When a gas dissolves in a dilute liquid-water drop and does not react chemically,
the quantities of the gas remaining and dissolved in water are governed by Henrys
law, which states that, over a dilute solution, the pressure exerted by a gas is
proportional to the molality of the dissolved gas in solution (Section 17.4). When
Henrys law holds, the quantity of gas over a drop surface is the saturation vapor
pressure of the gas, and the gas is in equilibrium with its dissolved phase at the
surface. As such, in the absence of aqueous chemistry, only a nite amount of a gas
dissolves in a liquid drop.
645
Table 19.1 Names, formulae, and lewis structures of S(IV) and S(VI) species
S(IV) family S(VI) family
Chemical name and Chemical Chemical name and Chemical
formula structure formula structure
Sulfur dioxide (g,aq)
SO
2
(g,aq)
S
O
O
Sulfurous acid (aq)
H
2
SO
3
(aq)
O
S
HO OH
Sulfuric acid (aq)
H
2
SO
4
(aq)
S
O
OH
O OH
Bisulte ion HSO

3
O
S
HO O
Bisulfate ion HSO

4
S
O
OH
O O
Sulte ion SO
2
3
O
S
O O
Sulfate ion SO
2
4
S
O
O
O O
When irreversible chemistry occurs, the dissolved gas chemically reacts to pro-
duce a product, instantaneously decreasing the molality of the dissolved gas. As
such, more gas must dissolve into the drop to maintain saturation over the drop
surface (i.e., to maintain the Henrys lawrelation). Once additional gas dissolves, it
reacts away irreversibly as well. The continuous dissolution and irreversible reac-
tion that occur in a drop can deplete the gas phase of dissolving species, sometimes
within minutes. Below, aqueous chemistry processes of importance are discussed.
19.2 MECHANISMS OF CONVERTING S(IV) TO S(VI)
The most important processes affected by irreversible chemistry are those that
convert sulfur dioxide gas to dissolved sulfuric acid and its dissociation products,
bisulfate and sulfate. These processes are important because sulfuric acid is the
most abundant acid in the air and the main culprit in acid deposition. Much of its
atmospheric production is due to irreversible chemistry within clouds, precipita-
tion, and aerosol particles.
Sulfur dioxide is a member of the S(IV) family. Dissolved sulfuric acid and its
products are members of the S(VI) family (Table 19.1). In these families, the IV
and the VI represent the oxidation states (+4 and +6, respectively) of the members
of the respective families. Thus, S(VI) members are more oxidized than are S(IV)
members. Because sulfur dioxide is in the S(IV) family and sulfuric acid, the main
source of acidity in rainwater, is in the S(VI) family, the oxidation of sulfur dioxide
646
19.2 Mechanisms of converting S(IV) to S(VI)
gas to aqueous sulfuric acid represents a conversion from the S(IV) family to the
S(VI) family. This conversion occurs through two pathways, described next.
19.2.1 Gas-phase oxidation of S(IV)
The rst conversion mechanism of S(IV) to S(VI) involves three steps. The initial
step is gas-phase oxidation of sulfur dioxide to sulfuric acid by
+OH(g), M +H
2
O(g)
Sulfur
dioxide gas
Bisulfite Sulfur
trioxide
Sulfuric
acid gas
+O
2
(g)
HO
2
(g)
SO
2
(g) HSO
3
(g) SO
3
(g) H
2
SO
4
(g)
(19.1)
The second step is condensation of sulfuric acid gas onto aerosol particles and
simultaneous hydration (Section 17.8) of liquid water to the sulfuric acid to produce
a sulfuric acidwater solution. Because sulfuric acid has a low saturation vapor
pressure, nearly all of it produced by (19.1) condenses, and this process is relatively
irreversible. The third step is dissociation of aqueous sulfuric acid. The second and
third steps are represented by
H
2
SO
4
(g) H
2
SO
4
(aq)

H
+
+HSO

4

2H
+
+ SO
2
4
Sulfuric Aqueous Bisulfate ion Sulfate ion
acid gas sulfuric acid
(19.2)
At typical pHs of aerosol particles and cloud drops, nearly all aqueous sulfuric acid
dissociates to the sulfate ion. The dissociation releases two protons, decreasing pH
and increasing acidity of the solution.
Condensation accounts for much of the increase of S(VI) in particles when the
relative humidity is below 100 percent. Below 70-percent relative humidity, nearly
all S(VI) production in particles is due to condensation. Because the rate of change
of mass of a particle due to condensation is proportional to the radius of the particle,
as shown in (16.13), condensation is a radius-limited process. Whereas gas-phase
oxidation followed by condensation is the dominant mechanism by which S(IV)
produces S(VI) in aerosol particles, particularly at low relative humidity, a second
mechanismproduces S(VI) fromS(IV) more rapidly in cloud drops and raindrops.
19.2.2 Aqueous-phase oxidation of S(IV)
The second conversion process of S(IV) to S(VI) involves several steps as well. The
initial step is reversible dissolution of sulfur dioxide gas into liquid drops by
SO
2
(g)

SO
2
(aq)
Sulfur Dissolved
dioxide gas sulfur dioxide
(19.3)
Sulfur dioxide is less soluble in water than is hydrochloric acid or nitric acid but
more soluble than are many other gases. Because the dissolution rate of sulfur
647
dioxide or any other gas depends on the volume of liquid water present, dissolution
is a volume-limited process.
The second step is reversible reaction of aqueous sulfur dioxide to sulfurous
acid, followed by dissociation of sulfurous acid:
SO
2
(aq) + H
2
O(aq)

H
2
SO
3
(aq)

H
+
+ HSO

3

2H
+
+ SO
2
3
Dissolved Liquid Sulfurous Hydrogen Bisulte Hydrogen Sulte
sulfur water acid ion ion ion ion
dioxide
(19.4)
At pH of 27, the major dissociation product of sulfurous acid is the bisulte ion
(HSO
3
). All sulfur species in (19.4) are in the S(IV) family, so the bisulte ion is
the major species present in the S(IV) family upon sulfur dioxide dissolution and
dissociation.
The third step is irreversible aqueous oxidation of S(IV) to S(VI). Over a range
of liquid-water contents (50 g m
3
to 2 g m
3
), hydrogen peroxide and ozone
are the major aqueous-phase oxidants of S(IV) at pH 6 and pH > 6, respec-
tively. Since oxidation of SO
2
by O
3
reduces pH, H
2
O
2
often becomes the most
important oxidant of SO
2
even when the initial pH > 6 (e.g., Liang and Jacobson
1999).
When the pH 6, hydrogen peroxide oxidizes S(IV) by
HSO

3
+ H
2
O
2
(aq) + H
+
SO
2
4
+ H
2
O(aq) +2H
+
Bisulte Hydrogen Sulfate
ion peroxide(aq) ion
(19.5)
This reaction is written in terms of the bisulte ion and the sulfate ion because, at
pH of 26, these are the two most prevalent components of the S(IV) and S(VI)
families, respectively. This reaction consumes S(IV) within tens of minutes if the
H
2
O
2
(aq) concentration exceeds that of S(IV). If the S(IV) concentration exceeds
that of H
2
O
2
(aq), then H
2
O
2
(aq) is depleted within minutes (Jacob 1986). The
rate coefcient of (19.5) multiplied by [H
+
] is 9 times greater at pH = 0 than at
pH=4 (Appendix Table B.8), suggesting that the coefcient increases with decreas-
ing pH. A cloud drop typically has a pH of 2.5 to 5.6. An aerosol particle often has
a pH of 1 to +2. The lower pH in aerosol particles speeds reaction rates in such
particles relative to those in cloud drops. But, because aerosol particles contain
less H
2
O
2
(aq) and liquid water than do cloud drops, aerosol particles are still less
efcient at converting S(IV) than are cloud drops.
Hydrogen peroxide is also photolyzed within an aqueous solution by
H
2
O
2
(aq) +h 2
OH(aq) (19.6)
648
and is lost by reaction with the aqueous hydroxyl radical,
H
2
O
2
(aq) +
OH(aq) H
2
O(aq) +H
O
2
(aq) (19.7)
Sources of hydrogen peroxide in liquid water are its dissolution by
H
2
O
2
(g)

H
2
O
2
(aq) (19.8)
and aqueous production from the hydroperoxy radical,
H
O
2
(aq) + O

2
+ H
2
O(aq) H
2
O
2
(aq) + O
2
(aq) + OH
Hydroperoxy Peroxy Hydrogen Hydroxide

radical ion peroxide ion
(19.9)
The aqueous hydroperoxy radical dissolves from the gas phase. At a pH
above 5, much of it dissociates. Dissolution and dissociation are represented
by
H
O
2
(g)

O
2
(aq)

H
+
+O

2
(19.10)
When the pH > 6, which occurs only in cloud drops that contain basic sub-
stances, such as ammonium or sodium, the most important reaction converting
S(IV) to S(VI) is
SO
2
3
+ O
3
(aq) SO
2
4
+ O
2
(aq)
Sulte Dissolved Sulfate Dissolved
ion ozone ion oxygen
(19.11)
This reaction is written in terms of the sulte ion and the sulfate ion because the
HSO
3
O
3
reaction is relatively slow, and at pH levels greater than 6, most S(VI)
exists as SO
4
2
. This reaction not only oxidizes S(IV), but also helps to deplete
gas-phase ozone in a cloud (e.g., Liang and Jacob 1997; Walcek et al. 1997; Zhang
et al. 1998). As the reaction proceeds, more ozone gas must dissolve to maintain
saturation over a drop surface.
When the liquid-water content exceeds 0.2 g m
3
and the pH 5, or when
H
2
O
2
is depleted, OH can also oxidize S(IV) by
HSO

3
+
OH(aq) + O
2
(aq) S
O

5
+H
2
O(aq)
Bisulte Peroxysulfate
ion ion
(19.12)
The hydroxyl radical is not very soluble in water, so its production from aqueous
chemistry generally exceeds its production from gas dissolution. The peroxysulfate
649
ion catalyzes oxidation of S(IV) by
HSO

3
+ O
2
(aq)
S
O

5
HSO

5
Bisulte Dissolved Peroxymonosulfate
ion oxygen ion
(19.13)
The peroxymonosulfate ion produced from this reaction also oxidizes S(IV) by
HSO

3
+ HSO

5
+H
+
2SO
2
4
+ 3H
+
Bisulte Sulfate
ion ion
(19.14)
Dissolved oxygen oxidizes S(IV) when a transition metal, such as Fe(III) or
Mn(II), catalyzes the reaction (Jacob et al. 1989a). About 8590 percent of soluble
iron appears as Fe(II) in the form of Fe
2+
, and about 1015 percent appears as
Fe(III), although these values are location-dependent. At pH - 12, most Fe(III) is
present as Fe
3+
and catalyzes S(IV) oxidation by
SO
2
3
+ H
2
O(aq) +O
2
(aq)
Fe(III)
SO
2
4
+ H
2
O
2
(aq)
Sulte ion Sulfate ion
(19.15)
Fe(II) does not catalyze S(IV) oxidation. Mn(II) is usually in the form of Mn
2+
and
catalyzes S(IV) by
HSO

3
+ H
2
O(aq) + O
2
(aq)
Mn(II)
SO
2
4
+ H
2
O
2
(aq) +H
+
Bisulte ion Sulfate ion
(19.16)
Mn(III) does not catalyze S(IV) oxidation signicantly.
Slightly soluble carbonyl compounds, such as formaldehyde and acetaldehyde,
also affect S(IV) oxidation in clouds and rain (e.g., Munger et al. 1989). These
species originate from the gas phase, where they are produced by combustion
and photochemistry. Dissolved formaldehyde reacts reversibly in solution to form
methylene glycol by
HCHO(aq) + H
2
O(aq)

H
2
C(OH)
2
(aq)
Formaldehyde Methylene glycol
(19.17)
Methylene glycol is lost during the reaction,
H
2
C(OH)
2
(aq) +
OH(aq) + O
2
(aq) HCOOH(aq) + H
O
2
(aq) +H
2
O(aq)
Methylene glycol Formic acid
(19.18)
650
At high pH, formaldehyde oxidizes the sulte ion to form hydroxymethanesul-
fonate (HMSA) by
SO
2
3
+ HCHO(aq) + H
2
O(aq) HOCH
2
SO

3
+OH
Sulte Formal HMSA

ion -dehyde
(19.19)
which is subsequently lost during the reaction
HOCH
2
SO

3
+
OH(aq) + O
2
(aq) HCHO(aq) +S
O

5
+H
2
O(aq)
HMSA Peroxysulfate
ion
(19.20)
Other slightly soluble organic gases include the methylperoxy radical (CH
3
O
2
),
methyl hydroperoxide (CH
3
OOH), formic acid (HCOOH), acetic acid
(CH
3
COOH), and peroxyacetic acid (CH
3
C(O)OOH). Relevant aqueous reac-
tions for these species appear in Appendix Table B.8, and equilibrium reactions
appear in Appendix Table B.7.
Other soluble inorganic gases include nitric acid, ammonia, and hydrochloric
acid. Under most conditions, the dissolved products of nitric acid and ammonia,
HNO
3
(aq), NO
3
, NH
3
(aq), and NH
4
+
are unaffected signicantly by irreversible
reactions (Pandis and Seinfeld 1989). The chemistry of these species is governed by
reversible reactions, discussed in Chapter 18. The chloride ion (Cl
), the dissoci-
ation product of HCl(aq), reacts irreversibly in solution. It maintains equilibrium
with several aqueous chlorine species through
Cl

2

Cl(aq) + Cl
Dichloride Chlorine Chloride

ion atom ion
(19.21)
ClOH
Cl
OH(aq)
Chlorine Chloride
hydroxide radical ion
(19.22)
(Jayson et al. 1973). The chlorine atom reacts irreversibly with hydrogen peroxide
by
H
2
O
2
(aq) +
Cl(aq) Cl
HO
2
(aq) +H
+
(19.23)
and the dichloride ion oxidizes S(IV) by
HSO

3
+ Cl

2
+ O
2
(aq) S
O

5
+2Cl
+H
+
Bisulte Dichloride Peroxysulfate
ion ion ion
(19.24)
but this reaction is less important than other reactions that oxidize S(IV).
In sum, the second conversion process of S(IV) to S(VI) involves dissolution
of sulfur dioxide gas to sulfurous acid, reversible dissociation of sulfurous acid
651
to the bisulte ion and the sulte ion, and irreversible oxidation of these ions to
the sulfate ion. Oxidants of bisulte and/or sulte in solution include hydrogen
peroxide, ozone, the hydroxyl radical, the peroxymonosulfate ion, formaldehyde,
the dichloride ion, and molecular oxygen catalyzed by iron, manganese, and the
peroxysulfate ion (e.g., Hoffmann and Calvert 1985; Jacob et al. 1989a; Pandis
and Seinfeld 1989).
The conversion rate of S(IV) to S(VI) depends on liquid-water content, oxi-
dant concentration, and pH. Aerosol liquid-water contents are generally less than
500 g m
3
. Fog liquid-water contents are generally less than 0.2 g m
3
. Cloud
liquid-water contents, which range from 0.22 g m
3
, exceed those of fogs. Thus,
a cloud has 50010 000 times more liquid water than does an aerosol plume.
The Henrys law constants of H
2
O
2
, HO
2
, OH, and O
3
are moderate to low,
as shown in Appendix Table B.7. Because of its high liquid-water content, a cloud
can absorb all four gases. Because of its lowliquid-water content, an aerosol plume
absorbs proportionately less of all four gases and absorbs only H
2
O
2
in signicant
quantities. Thus, S(IV) conversion in an aerosol plume is often limited by the
concentration of H
2
O
2
. In a cloud, the greater abundance of liquid water, H
2
O
2
,
HO
2
, OH, and O
3
in solution permits S(IV) to convert to S(VI) faster than in
an aerosol plume. Although S(IV) oxidation is slow in aerosol particles, it can be
enhanced at low temperatures (268 K) if the relative humidity is high.
The faster that S(IV) oxidation occurs, the faster that sulfur dioxide is depleted
from the gas phase, and the faster that S(VI) forms. In cloud drops, dissolution
and aqueous reaction can convert 60 percent of gas-phase sulfur dioxide to the
sulfate ion within 20 minutes (e.g., Jacobson 1997a; Liang and Jacobson 1999).
Figure 19.1 illustrates this point. The gure compares modeled in-cloud concen-
trations of liquid water, the bisulfate ion, and the sulfate ion over time when
sulfur dioxide dissolution and aqueous reaction were and were not included in
the calculation. Within 20 min, cloud drops dissolved and converted more than
30 g m
3
(58 percent) out of an initial 52 g m
3
of SO
2
(g). Aerosol particles,
on the other hand, require hours to days to convert the same quantity of sulfur
dioxide.
Conversion of S(IV) to S(VI) in aerosol particles, cloud drops, and precipitation
drops contributes to acid deposition. Acid deposition occurs to some extent in
and downwind of all countries worldwide. It is particularly prevalent in provinces
of eastern Canada, the northeastern United States (particularly the Adirondack
Mountain region), southern Scandinavia, middle and eastern Europe, India, Korea,
Russia, China, Japan, Thailand, and South Africa. The natural pH of cloud water
and rainwater is 5.6 (Fig. 17.1). Aqueous oxidation of S(IV) to S(VI) generally
reduces the pH in fog and rain to 25.6, creating acid deposition.
19.3 DIFFUSION WITHIN A DROP
When a gas dissolves in a liquid drop, it immediately begins to diffuse through the
drop and reacts chemically at the same time. The characteristic time scale (s) for
652
19.3 Diffusion within a drop
0
5
10
15
20
25
30
35
40
0 0.5 1 1.5 2
S
u
m
m
e
d

c
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Time from start (h)
H
2
O 0.0001 (a and b)
SO
4
2
(a)
HSO
4
(b)
HSO
3
(b)
HSO
4
(a)
SO
4
2
(b)
Figure 19.1 Time-series comparison of dissolved

ion and water concentrations, summed over all
cloud drop sizes, when cloud drops grew and sul-
fur dioxide (a) did not and (b) did dissolve in and
react within the drops. H
2
SO
4
condensed in both
cases, but SO
2
dissolved and reacted in the second
case only. Initial conditions were SO
2
(g), 52 g m
3
;
H
2
O
2
(g), 10 g m
3
; HNO
3
(g), 30 g m
3
; NH
3
(g),
10 g m
3
; HCl(g), 0 g m
3
; H
2
SO
4
(g), 15 g m
3
;
H
2
SO
4
(aq), 10 g m
3
; NaCl(aq), 15 g m
3
; and
T = 298 K. The cloud was formed by increasing the
relative humidity to 100.01 percent at the start, solv-
ing the growth equations with a 1-s time step, and
replenishing the relative humidity to 100.01 percent
every time step for the rst 10 min of simulation.
After 10 min, the relative humidity was allowed to
relax to its calculated value. FromJacobson (1997a).
diffusion of solute q through a liquid drop is
ad.q
=
r
2
i
2
D
aq.q
(19.25)
(Crank 1975) where D
aq.q
is the diffusion coefcient of the solute in water at
25

C (cm
2
s
1
) and r
i
is the radius of the drop (cm). A typical diffusion coefcient
is D
aq.q
=2 10
5
cm
2
s
1
(e.g., Wilke and Chang 1955; Himmelblau 1964; Jacob
1986; Schwartz 1986).
Example 19.1
From (19.25), the characteristic time for diffusion within a 30-m-diameter
cloud drop is about 0.011 s. The characteristic time for diffusion within a
10-m-diameter drop is about 0.0013 s. Thus, for many purposes, diffusion
can be assumed instantaneous.
653
Although the time scale for diffusion is small, chemical reactions often occur
on an even smaller time scale. Jacob (1986) found that O
3
(aq), NO
3
(aq), OH(aq),
Cl(aq), SO

4
, CO

3
, and Cl

2
had chemical lifetimes shorter than diffusion trans-
port times. Such species react near the surface of a drop before they can diffuse to
the center. When diffusion and chemical reaction are considered, the time rate of
change of concentration of species q in size bin i as a function of radius r is
dc
q.i.r
dt
ad.aq
= D
aq.q
1
r
2
r
2
c
q.i.r
r
+ P
c.q.i.r
L
c.q.i.r
(19.26)
where P
c.q.i.r
and L
c.q.i.r
are the net aqueous chemical production and loss rates
(mol cm
3
s
1
), respectively, of q in size bin i at a distance r from the center of
the drop, and the subscript ad, aq indicates that aqueous diffusion is coupled with
aqueous chemistry. At the drop center, c
q.i.r
r = 0.
19.4 SOLVING GROWTH AND AQUEOUS CHEMICAL ODES
Aqueous chemical reactions are similar to gas-phase reactions in that both are
described by rst-order ODEs that can be solved with the methods discussed in
Chapter 12. Aqueous chemistry ODEs are more difcult to solve than gas chem-
istry ODEs for two reasons. First, the e-folding lifetimes of aqueous species against
chemical loss are usually shorter than are those of gases. Thus, aqueous reactions
are stiffer than gas reactions. Second, when dissolved gases are destroyed by aque-
ous reactions, the dissolved gases are rapidly replenished by additional dissolution
to maintain saturation over the particle surface. When H
2
O
2
(aq) reacts in solution,
for example, H
2
O
2
(g) dissolves to replace lost H
2
O
2
(aq) to satisfy (19.8). Thus, in
a model that resolves aerosol particles and hydrometeor particles among multiple
size bins, aqueous ODEs must be solved within each size bin, and growth ODEs
must be solved between the gas phase and all bins simultaneously. Gas chemistry
is effectively solved in one size bin.
To make matters more complex, many dissolving gases dissociate reversibly in
solution. Thus, dissolution, equilibrium, and aqueous reaction processes must be
solved nearly simultaneously. The rate of change of particle concentration due to
these processes is
dc
q.i.t
dt
ge.eq.aq
=
dc
q.i.t
dt
ge
+
dc
q.i.t
dt
eq
+
dc
q.i.t
dt
aq
(19.27)
where c
q.i.t
is the mole concentration of species q in bin i at time t, and subscripts
ge, eq, and aq identify growth/evaporation, equilibrium, and aqueous chemistry,
respectively. The aqueous chemistry term in (19.27) is the same as that in (19.26),
except that diffusion is assumed to be instantaneous and particles are assumed to be
well mixed in (19.27). The gas-aqueous mole conservation equation corresponding
to (19.27) is
dC
q.t
dt
=
N
B
i =1
dc
q.i.t
dt
ge
(19.28)
654
19.4 Solving growth and aqueous chemical ODEs
where C
q.t
is the gas mole concentration of species q, and N
B
is the number of size
bins. Together, (19.27) and (19.28) make up a large set of stiff ODEs.
An exact solution to (19.27) and (19.28) is difcult to obtain. The equilibrium
term is not an ordinary differential equation. It is a reversible equation solved with
an iterative method, such as the MFI or AEI schemes discussed in Chapter 17. A
reversible equation, of the formD+E
A +B, can be written as a combination of

two ODEs, D+E A +B and A +B D+E. These equations are very stiff and
time-consuming to solve. Instead, the reaction D+E
A +B is usually simulated
as a reversible equation that may be operator-split from the other terms in (19.27).
The growth and aqueous-chemistry terms in (19.27) are often solved together
in zero- and one-dimensional modeling studies (e.g., Chameides 1984; Schwartz
1984; Jacob 1986; Pandis and Seinfeld 1989; Bott and Carmichael 1993; Sander
et al. 1995; Zhang et al. 1998). In some cases, the equations are solved with an inte-
grator of stiff ODEs. In others, they are solved with a forward Euler method taking
a small time step. Here, a method of solving the equations in a three-dimensional
model is discussed.
With this method, species involved in dissociation/association reactions among
themselves are grouped in families. Family species cycle rapidly among each
other due to rapid and reversible dissociation/association. Changes in family con-
centrations are slower than changes in species concentrations. Here, the S(IV),
S(VI), HO
2.T
,CO
2.T
. HCHO
T
. HCOOH
T
, and CH
3
COOH
T
families are consid-
ered. Family mole concentrations (moles per cubic centimeter of air) in size bin
i are
c
S(IV).i
= c
SO
2
(aq).i
+c
HSO
3
.i
+c
SO
3
2
.i
(19.29)
c
S(VI).i
= c
H
2
SO
4
(aq).i
+c
HSO
4
.i
+c
SO
4
2
.i
(19.30)
c
HO
2.T
.i
= c
HO
2
(aq).i
+c
O
2
.i
(19.31)
c
CO
2.T
.i
= c
CO
2
(aq).i
+c
HCO
3
.i
+c
CO
3
2
.i
(19.32)
c
HCHO
T
.i
= c
HCHO(aq).i
+c
H
2
C(OH)
2
.i
(19.33)
c
HCOOH
T
.i
= c
HCOOH(aq).i
+c
HCOO
.i
(19.34)
c
CH
3
COOH
T
.i
= c
CH
3
COOH(aq).i
+c
CH
3
COO
.i
(19.35)
respectively. Some species, such as H
2
O
2
(aq) and O
3
(aq), are each self-contained
in their own family since H
2
O
2
(aq) dissociates signicantly only at a pH higher
than observed in cloud drops, and O
3
(aq) does not dissociate in solution. Growth
aqueous-chemical ODEs can be obtained for families by combining (17.99) with
chemical production and loss terms. For the S(IV) family, the resulting ODE for a
size bin is
dc
S(IV).i
dt
= k
S(IV).i
C
SO
2
(g)
S
S(IV).i
c
SO
2
(aq).i
H
SO
2
(aq).i
+ P
c.S(IV).i
L
c.S(IV).i
(19.36)
where k
S(IV).i
is the coefcient for mass transfer of SO
2
to particles of size i (s
1
) from
(16.64), C
SO
2
(g)
is the concentration of SO
2
(g) (mol cm
3
), S
S(IV).i
is the equilibrium
saturation ratio from (16.47), P
c.S(IV).i
is the aqueous chemical production rate of
655
S(IV) in size bin i (mol cm
3
s
1
), L
c.S(IV).i
is the aqueous chemical loss rate of S(IV)
in bin i (mol cm
3
s
1
), and
H
SO
2
(aq).i
= m
v
c
w
R
TH
SO
2
(19.37)
is the dimensionless Henrys constant of SO
2
. Parameters in (19.37) were dened
in Section 17.14.1. The concentrations c
SO
2
(aq).i
and c
S(IV).i
are related by
c
SO
2
(aq).i
= c
S(IV).i
m
2
H
+
.i
m
2
H
+
.i
+m
H
+
.i
K
1.S(IV)
+ K
1.S(IV)
K
2.S(IV)
(19.38)
derived from (19.29), (17.97), and
K
1.S(IV)
=
m
H
+
.i
m
HSO
3
.i
2
i .H
+
HSO
3
m
SO
2
(aq).i
(19.39)
K
2.S(IV)
=
m
H
+
.i
m
SO
3
2
.i
2
i .H
+
SO
3
2
m
HSO
3
.i
2
i .H
+
HSO
3
(19.40)
which are the rst and second equilibrium dissociation coefcient expressions for
the reversible reactions,
SO
2
(aq) +H
2
O

H
+
+HSO

3
(19.41)
HSO
3
H
+
+SO
2
3
(19.42)
respectively. For simplicity, the activity coefcients in (19.39) and (19.40) may be
set to zero when dilute solutions (e.g., cloud drops) are considered.
Substituting (19.38) into (19.36) and adding time subscripts give
dc
S(IV).i.t
dt
= k
S(IV).i.th
C
SO
2
(g).t
S
S(IV).i.th
c
S(IV).i.t
H
S(IV).i.th
+ P
c.S(IV).i.t
L
c.S(IV).i.t
(19.43)
where
H
S(IV).i.th
= m
v
c
w.i.th
R
TH
SO
2
1 +
K
1.S(IV)
m
H
+
.i.th
+
K
1.S(IV)
K
2.S(IV)
m
2
H
+
.i.th
(19.44)
is the dimensionless effective Henrys constant of S(IV).
The chemical production and loss rates of a species q are determined from
P
c.q.i.t
=
N
prod.q
l=1
R
c.n
P
(l.q).t
L
c.q.i.t
=
N
loss.q
l=1
R
c.n
L
(l.q).t
(19.45)
where N
prod.q
(N
loss.q
) is the number of reactions in which species q is produced
(lost), R
c.n
P
(l.q)
[R
c.n
L
(l.q)
] is the lth production (loss) rate of q, and n
P
(l. q) [n
L
(l. q)]
656
19.4 Solving growth and aqueous chemical ODEs
gives the reaction number corresponding to the lth production (loss) termof species
q. The terms in (19.44) are similar to those in (12.53) and (12.55), respectively.
S(IV) has no important aqueous chemical sources but several sinks.
Appendix Table B.8 lists S(IV) oxidation reactions. Two of the reactions
S(IV) +H
2
O
2
(aq) +H
+
S(VI) +2H
+
+H
2
O(aq) (19.46)
S(IV) +HO
2.T
S(VI) +OH(aq) +2H
+
(19.47)
are considered here to illustrate the calculation of the chemical loss rate of the
S(IV) family for use in a chemical solver. Appendix Table B.8 shows that the only
important S(IV) reactant in (19.46) is HSO
3
:
HSO
3
+H
2
O
2
(aq) +H
+
(19.48)
In (19.47), HSO
3
and SO
3
2
are important S(IV) reactants and HO
2
(aq) and
O
2
are important HO
2,T
reactants. As such, (19.47) is really a combination of the
following four reactions:
HSO
3
+ HO
2
(aq) (19.49)
SO
3
2
+ HO
2
(aq) (19.50)
HSO
3
+ O
2
(19.51)
SO
3
2
+ O
2
(19.52)
One way to consider the loss of S(IV) in a chemical solver is to treat reactions
(19.48)(19.52) as separate reactions. Doing so requires the simultaneous solution
of the equilibrium equations
SO
2
(aq) +H
2
O(aq)

H
+
+HSO
3
2H
+
+SO
3
2
(19.53)
HO
2
(aq)

H
+
+O
2
(19.54)
Alternatively, the chemical loss of S(IV) or any other species can be written in
terms of the family reactions, (19.46) and (19.47), where the families are dened
in (19.29)(19.35). In this case, the instantaneous equilibrium partitioning that
occurs with (19.53) and (19.54) is accounted for in the loss rate coefcient. For
example, in the case of S(IV) reacting in (19.46) and (19.47) only, the chemical loss
rate of S(IV) is
L
c.S(IV).i.t
= k
a
c
S(IV).i.t
c
H
2
O
2
.i.t
c
H
+
.i.t
+k
b
c
S(IV).i.t
c
HO
2.T
.i.t
(19.55)
(mol cm
3
s
1
). In this equation,
k
a
= k
a.1
1.S(IV)
(19.56)
657
is a third-order rate coefcient (cm
6
mol
2
s
1
), where k
a.1
is the rate coefcient
for Reaction (19.48), given in Appendix Table B.8 (converted from (M
2
s
1
) to
(cm
6
mol
2
s
1
)), and
1.S(IV)
=
m
H
+
.i.th
K
1.S(IV)
m
2
H
+
.i.th
+m
H
+
.i.th
K
1.S(IV)
+ K
1.S(IV)
K
2.S(IV)
(19.57)
is the mole fraction of total S(IV) partitioned to HSO
3
. This fraction depends

on pH, so m
H
+ must be solved for simultaneously. The equilibrium coefcients in
(19.57) are given in (19.39) and (19.40).
The second rate coefcient in (19.55) is
k
b
=
k
b.1
1.S(IV)
+k
b.2
2.S(IV)
0.HO
2.T
+
k
b.3
1.S(IV)
+k
b.4
2.S(IV)
1.HO
2.T
(19.58)
(cm
3
mol
1
s
1
) where k
b.1
, k
b.2
, k
b.3
, and k
b.4
are the rate coefcients from
Appendix Table B.8 for Reactions (19.49)(19.52), respectively,
2.S(IV)
=
K
1.S(IV)
K
2.S(IV)
m
2
H
+
.i.th
+m
H
+
.i.th
K
1.S(IV)
+ K
1.S(IV)
K
2.S(IV)
(19.59)
is the mole fraction of total S(IV) partitioned to SO
3
2
, and
0.HO
2.T
=
m
H
+
.i.th
m
H
+
.i.th
+ K
1.HO
2.T
(19.60)
1.HO
2.T
=
K
1.HO
2.T
m
H
+
.i.th
+ K
1.HO
2.T
(19.61)
are the mole fractions of HO
2,T
partitioned to HO
2
(aq) and O
2
, respectively.
In these expressions, K
1.HO
2.T
is the equilibrium coefcient for Reaction (19.54).
The family method described above signicantly reduces the number of reactions
required for a solution to aqueous chemical equations.
The gas conservation equation corresponding to (19.43) is
dC
SO
2
(g).t
dt
=
N
B
i =1
k
S(IV).i.th
C
SO
2
(g).t
S
S(IV).i.th
c
S(IV).i.t
H
S(IV).i.th
(19.62)
Together, (19.43) and (19.62) make up a set of ODEs whose order depends on
the number of dissolving gases, particle size bins, and species taking part in chem-
ical reactions. This set of equations can be solved in a three-dimensional model
with a sparse-matrix solver. Table 19.2 shows that such a solver reduced the num-
ber of multiplications during one call to a matrix decomposition subroutine by
over two million (99.7 percent) for a given set of species, reactions, and size
bins.
The last four rows showthe number of operations in each of four loops of matrix
decomposition and backsubstitution. The percentages are values in the second
658
19.6 Problems
Table 19.2 Reduction in array space and number of matrix operations
before and after the use of sparse-matrix techniques for ODEs arising
from 10 gases transferring to 16 size bins, and 11 aqueous chemistry
reactions occurring among 11 species within each bin
Initial After sparse reductions
Order of matrix 186 186
No. init. matrix positions lled 34 596 1226 (3.5%)
No. nal matrix positions lled 34 596 2164 (6.3%)
No. operations decomp. 1 2 127 685 6333 (0.3%)
No. operations decomp. 2 17 205 1005 (5.8%)
No. operations backsub. 1 17 205 1005 (5.8%)
No. operations backsub. 2 17 205 973 (5.6%)
column relative to those in the rst. The solver used was SMVGEAR II (Jacobson
1998a).
19.5 SUMMARY
In this chapter, aqueous chemical reactions and their effects on aerosol and hydrom-
eteor composition were discussed. Aqueous reactions are important because they
affect the rate of chemical conversion of sulfur dioxide gas, a member of the
S(IV) family, to the sulfate ion, a member of the S(VI) family. Production of the
sulfate ion in a liquid drop increases the acidity of the drop. Thus, conversion
of S(IV) to S(VI) in atmospheric particles enhances acid deposition. The rate of
S(IV) conversion to S(VI) depends on liquid-water content, S(IV) oxidant con-
centrations, and pH. Conversion is faster in cloud drops than in aerosol parti-
cles because clouds have higher liquid-water contents than do aerosol particles.
Numerical methods of solving aqueous chemistry ODEs are the same as those for
solving gas-chemistry ODEs. Aqueous ODEs are more difcult to solve because
they are stiffer and more tightly coupled to other processes than are gas-phase
ODEs.
19.6 PROBLEMS
19.1 What are the two primary formation mechanisms of S(VI) in particles? How
might changes in relative humidity affect the rate of growth of particles due
to each mechanism?
19.2 Why do aerosol particles convert S(IV) species to S(VI) species at a lower
rate than do cloud drops?
19.3 Draw approximate curves of relative concentration versus radius when the
time scale for diffusion in a drop is (a) less than, (b) equal to, (c) greater than
that for chemical reaction loss of an aqueous species. Assume the relative
concentration at the drop center is unity.
19.4 Why are aqueous-phase reactions generally stiffer than gas-phase reactions?
659
19.5 Calculate the e-folding lifetimes of HSO
3
against loss by reaction with

O
3
(aq), H
2
O
2
(aq), and OH(aq), respectively. Assume
O
3(
g
)
,
H
2
O
2(
g
)
, and
OH
(
g
)
are 0.1, 8 10
4
, and 4.0 10
8
ppmv, respectively. Estimate aqueous
molalities of these species using Henrys constant from Appendix Table B.7.
Assume that T = 280 K, p
d
= 950 hPa, the solution is dilute, and pH = 4.5.
19.6 Write a computer script to read reactions and rate coefcients. Use the script
to calculate rate coefcients for all reactions in Appendix Table B.8 when
T = 298 K and when 273 K.
660
20
Sedimentation, dry deposition, and airsea exchange
A
erosol particles fall in the atmosphere due to the force of gravity. Their fall
speeds are determined by a balance between the force of gravity and the oppos-
ing force of drag that arises due to air viscosity. The sinking of particles at their fall
speed is called sedimentation. Gases also sediment, but their weights are so small
that their fall speeds are negligible. A typical gas molecule has a diameter on the
order of 0.5 nm (nanometer). Such diameters result in sedimentation (fall) speeds
on the order of 10 cm per year. Because this speed is slow, and because the slightest
turbulence moves gas molecules much further in seconds than sedimentation moves
them in years, gas fall speeds are important only over time scales of thousands of
years. Small-particle sedimentation is also of limited importance in the atmosphere
over short periods, but sedimentation of large particles, particularly soil dust, sea
spray, cloud drops, and raindrops, is important. Dry deposition removes gases
and particles at airsurface interfaces. Dry deposition occurs when gases or par-
ticles contact a surface and stick to or react with the surface. Sedimentation is
one mechanism by which particles contact surfaces. Others, which affect gases and
particles, are molecular diffusion, turbulent diffusion, and advection. In this chap-
ter, equations for sedimentation and dry deposition are described and applied.
In addition, uxes of gases between the ocean and atmosphere are discussed.
Removal of gases and particles by precipitation (wet deposition) was outlined in
Chapter 18.
20.1 SEDIMENTATION
The fall speed of a particle through air is determined by equating the downward
force of gravity with the opposing drag force that arises from air viscosity, as
illustrated in Fig. 20.1.
When a particle falling through the air is larger than the mean free path of an
air molecule (about 63 nm at 288 K and 1013 hPa from (15.24)) but small enough
so that the inertial force the particle imparts is less than the viscous force exerted
by air, the ow of the particle through the air is called Stokes ow. During Stokes
ow, the drag force of air on a sphere is governed by Stokes law,
F
D
= 6r
i
a
V
f.i
(20.1)
where r
i
is the particle radius (cm),
a
is the dynamic viscosity of air (g cm
1
s
1
),
and V
f.i
is the sedimentation (fall) speed (cm s
1
). An expression for the dynamic
661
Drag
Gravity
Figure 20.1 Ver-
tical forces acting
on a particle.
viscosity of air in units of kg m
1
s
1
was given in (4.54). Equation (20.1) illustrates
that the drag force is proportional to the fall speed of the particle.
When a particle is smaller than the mean free path of an air molecule, motion
through air is called slip ow. For such ow, Stokes law can be modied with the
Cunningham slip-ow correction and KnudsenWeber term, dened in (15.30)
as G
i
= 1 +Kn
a.i
[A
+B
exp(C
Kn
1
a.i
)]. With this correction, the drag force
becomes
F
D
=
6r
i
a
V
f.i
G
i
=
6r
i
a
V
f.i
1 +Kn
a.i
+B
exp
Kn
1
a.i
(20.2)
The smaller the particle, the larger the Knudsen number for air, the larger G
i
, and
the smaller the force of drag on a falling sphere. For Stokes ow, which occurs for
relatively larger particles, G
i
approaches unity.
The downward force of gravity, net of buoyancy, acting on a spherical part-
icle is
F
G
=
4
3
r
3
i
(
p
a
)g (20.3)
where
p
is the density of the particle (g cm
3
),
a
is the density of air (g cm
3
),
and g is the gravitational acceleration (cm s
2
).
Equating drag and gravitational forces for a falling spherical particle and solving
for the fall speed give
V
est
f.i
=
2r
2
i
(
p
a
)g
9
a
G
i
(20.4)
In this equation, the superscript est indicates that the fall speed is an estimated
value, since this equation will be used shortly in a corrective procedure. When
G
i
>1, which occurs for relatively small particles in the slip ow regime, the fall
speed is enhanced due to the lowviscosity of the air. When G
i
=1, which occurs for
larger particles, (20.4) simplies to the Stokes terminal fall speed, which is accurate
for Stokes ow only.
662
20.1 Sedimentation
Example 20.1
Estimate the fall speed of a 1-m-diameter particle of density 1.5 g cm
3
in
the stratosphere, where T = 220 K and p
a
= 25 hPa. Assume the particle is
spherical and the air is dry. Howmuch does the fall speed change at a pressure
of 1013 hPa if all other conditions are the same?
SOLUTION
From (4.54), the dynamic viscosity of air is 1.436 10
4
g cm
1
s
1
. From
(2.3), the thermal speed of an air molecule is 4.01 10
4
cm s
1
. From (2.36),
the density of air is 3.96 10
5
g cm
3
. From (15.24), the mean free path of an
air molecule is 1.769 10
4
cm. From (15.24), the Knudsen number for air is
3.54. From (15.30), the Cunningham slip-ow correction is 6.58. Finally, from
(20.4), the estimated fall speed is
V
f.0.05m

2(0.00005 cm)
2
1.5 0.0000396
g
cm
3
980.7
cm
s
2
9 1.436 10
4
g
cms
6.58
= 0.037
cm
s
At an air pressure of 1013 hPa, the fall speed for this problemis 0.0063 cms
1
.
Thus, fall speed decreases with increasing pressure (decreasing altitude).
The fall speed is related to the particle Reynolds number, rewritten from (15.26)
as
Re
est
i
=
2r
i
V
est
f.i
a
(20.5)
Both slip ow and Stokes ow occur at Reynolds numbers -0.01, which corre-
spond to particles -20 min diameter under near-surface conditions. At Reynolds
numbers >0.01, particles are larger than those in the Stokes ow regime, and
another correction to the drag force is necessary.
Beard (1976) studied the fall speeds of cloud drops and raindrops in air and
found that the correction factor for such particles can be parameterized as a func-
tion of the physical properties of the drops and their surroundings. Such properties
include temperature, pressure, air density, air viscosity, particle density, surface ten-
sion, and gravity. Beards parameterization requires two steps. The rst step is to
estimate the fall speed and Reynolds number with (20.4) and (20.5), respectively.
The second step is to recalculate the Reynolds number for one of three owregimes:
slip ow around a rigid sphere at low Reynolds number (-0.01), continuum ow
past a rigid sphere at moderate Reynolds number (0.01300), and continuum ow
around a nonspherical, equilibrium-shaped drop at large Reynolds number (>300).
These three regimes correspond to particle diameters of -20 m (aerosol particles
and cloud drops), 20 m1 mm (cloud drops to medium raindrops), and 15 mm
663
Table 20.1 Coefcients for polynomial ts used in (20.6)
B
0
= 3.186 57 E
0
= 5.000 15
B
1
= 0.992 696 E
1
= 5.237 78
B
2
= 0.001 531 93 E
2
= 2.049 14
B
3
= 0.000 987 059 E
3
= 0.475 294
B
4
= 0.000 578 878 E
4
= 0.054 281 9
B
5
= 0.000 085 517 6 E
5
= 0.002 384 49
B
6
= 0.000 003 278 15
Source: Beard (1976).
(large raindrops), respectively. In sum,
Re
nal
i
=
2r
i
V
est
f.i

a
Re
est
i
- 0.01
G
i
exp(B
0
+ B
1
X+ B
2
X
2
+ ) 0.01 Re
est
i
- 300
N
16
P
G
i
exp(E
0
+ E
1
Y + E
2
Y
2
+ ) Re
est
i
300
(20.6)
where
X= ln
32r
3
i
(
p
a
)
a
g
3
2
a
Y = ln
4
3
N
Bo
N
1
/
6
P
(20.7)
are functions of the physical properties of particles and the atmosphere.
These equations contain the dimensionless physical-property number and Bond
number,
N
P
=
3
wa
2
a
4
a
(
p
a
)g
N
Bo
=
4r
2
i
(
p
a
)g
wa
(20.8)
respectively, where
wa
is the surface tension of water against air, dened in (14.19).
The Bond number measures the relative strength of the drag and surface tension
forces for a drop at terminal fall speed (Pruppacher and Klett 1997). The B and E
coefcients in (20.6) are given in Table 20.1. After the nal Reynolds number is
found, a nal fall speed is calculated with
V
f.i
= V
nal
f.i
=
Re
nal
i

a
2r
i
(20.9)
In the low-Reynolds-number regime (diameters -20 m), (20.9) simplies to the
Stokes terminal fall speed with the Cunninghamcorrection factor. In the moderate-
Reynolds-number regime (20 m1 mmdiameter), drops are essentially spherical,
and neither surface tension nor internal viscosity affects their shape or sedimenta-
tion speed signicantly. Large raindrops (>1 mm) are affected by turbulence and
vary in shape from oblate to prolate spheroids.
664
20.2 Dry deposition
Table 20.2 Time for a particle (or a gas for the smallest size) to fall
1 km in the atmosphere by sedimentation
Diameter (m) Time to fall 1 km Diameter (m) Time to fall 1 km
0.0005 9630 y 4 23 d
0.02 230 y 5 14.5 d
0.1 36 y 10 3.6 d
0.5 3.2 y 20 23 h
1 328 d 100 1.1 h
2 89 d 1000 4 m
3 41 d 5000 1.8 m
Results were obtained from the fall speeds in Fig. 20.3
Beards parameterization is valid for liquid water drops. Fall-speed parameter-
izations for ice crystals and aerosol particles of various shapes have also been
derived. Some shapes that have been considered include prolate ellipsoids, oblate
ellipsoids, circular cylinders, and rectangular parallelepipeds (Fuchs 1964); cubes,
hexagons, and cylinders (Turco et al. 1982); and columnar crystals, platelike crys-
tals, oblate spheroids, hexagonal crystals, hailstones, and graupel (Pruppacher and
Klett 1997).
The time for a particle to fall a distance Lz in the air due to sedimentation is
f.i
= LzV
f.i
. Table 20.2 shows the time required for particles to fall 1 km by
sedimentation under near-surface conditions. Emitted volcanic particles range in
diameter from -0.1 to >100 m. Particles 100 m in diameter take 1.1 h to fall
1 km. The only volcanic particles that survive more than a few months before
falling out of the atmosphere are those smaller than 4 m. Such particles require
no less than 23 days to fall 1 km. Most aerosol particles of atmospheric importance
are 0.11 min diameter. Table 20.2 indicates that these particles require 36 years
328 days, respectively, to drop 1 km. Gases, whose diameters are on the order of
0.5 nm (0.0005 m), require about 9600 years to fall 1 km.
Fall speeds are useful for calculating not only downward uxes of particles
but also coagulation kernels (Section 15.5) and dry-deposition speeds (discussed
next).
20.2 DRY DEPOSITION
Dry deposition occurs when a gas or particle is removed at an airsurface interface,
such as on the surface of a tree, a building, a window, grass, soil, snow, or water
(e.g., Sehmel 1980). A pollutant dry deposits when it impacts and sticks to the
surface. A deposited pollutant may be resuspended into the air when wind or an
eddy blows over a surface and lifts the pollutant.
Dry-deposition speeds are generally parameterized as the inverse sum of a series
of resistances (Wesely and Hicks 1977; Slinn et al. 1978; McRae et al. 1982; Walcek
et al. 1986; Russell et al. 1993). With the resistance model, the dry-deposition speed
665
b
s
a
r
z
R
R
R
z
R
R
R
Figure 20.2 Locations where gas dry-deposition
resistances apply. The gure is not to scale.
of a gas (m s
1
) is
V
d.gas
=
1
R
a
+ R
b
+ R
s
(20.10)
where R
a
is the aerodynamic resistance between a reference height (about 10 m
above the surface) and the laminar sublayer adjacent to the surface, R
b
is the
resistance to molecular diffusion through the 0.1 to 0.01-cm-thick laminar sublayer,
and R
s
is the resistance to chemical, biological, and physical interaction and sticking
between the surface and the gas once the gas has collided with the surface. Because
gases are so light and take so long to fall (Table 20.2), their dry-deposition speeds
do not need to include a sedimentation term. Resistances have units of time per
unit distance (e.g., s m
1
). Figure 20.2 depicts the region near the surface where
the three resistances listed apply.
The aerodynamic resistance is a property of the turbulent transfer between a
reference height and the laminar sublayer and depends on atmospheric stability.
The resistance to molecular diffusion depends on the gas diffusion coefcient, air
viscosity, air conductivity, air temperature, and air density. The resistance to surface
interactions depends on physical and chemical properties of the surface and the
depositing gas.
Dry-deposition speeds of particles differ from those of gases in two ways. First,
particles are heavier than gases; so, particle sedimentation speeds must be included
in the particle dry-deposition equation. Second, because of their weight, particles
tend to stay on a surface once they deposit (in the absence of strong winds or
turbulence creating resuspension); thus, the surface resistance termmay be ignored.
The dry-deposition speed of a particle (m s
1
) of size i is
V
d.part.i
=
1
R
a
+ R
b
+ R
a
R
b
V
f.i
+ V
f.i
(20.11)
where the fall speed is in meters per second.
666
20.2 Dry deposition
For gas and particle dry deposition, the aerodynamic resistance (s m
1
) can be
approximated from similarity theory with
R
a
=
z
r
z
0.q
h
dz
z
ku
(20.12)
where k is the dimensionless von K arm an constant (0.40), u
is the friction wind

speed (m s
1
), z is the height above the surface (m), z
r
is a reference height (10 m),
z
0,q
is the surface roughness length of gas or particle q (m), and
h
is a dimensionless
potential temperature gradient. A dimensionless potential temperature gradient is
used instead of a dimensionless wind shear because the transport of trace gases
resembles the transport of energy more than it does that of momentum. The friction
wind speed, dened in Chapter 8, is a measure of the vertical turbulent ux of
horizontal momentum. A large value of u
corresponds to high turbulence, low

resistance to vertical turbulent motion, and a lowdry deposition speed. The surface
roughness length of species q, (z
0.q
), can be found from (8.13) by substituting the
diffusion coefcient of the gas or particle of interest for that of water vapor.
Integrals of
h
were given in (8.38) as a function of the MoninObukhov length
L, which relies on u
and the turbulent temperature scale
. Anoniterative method
of determining u
, L, and
was discussed in Section 8.4.2.4. During the day,

winds are stronger, the atmosphere is more unstable, u
is larger, the heat integral

is smaller, and R
a
is smaller than during the night. Because resistance to turbulent
transfer is smaller during the day than night, dry-deposition speeds are generally
larger during the day than night.
The reciprocal of an aerodynamic resistance is a transfer speed, which gives the
rate at which a gradient in concentration of a species is transferred between z
0.q
and z
r
. Equation (8.53) gave the vertical turbulent ux of moisture (kg m
2
s
1
)
from MoninObukhov similarity theory as E
f
=
a
q
. Substituting (8.52) and

(20.12) into (8.53) and assuming species q is water vapor, give another expression
for the turbulent moisture ux, in terms of aerodynamic resistance, as
E
f
=
a
ku
[q
v
(z
r
) q
v
(z
0.v
)]
z
r
z
0.v
z
L
dz
z
=
a
R
a
[q
v
(z
r
) q
v
(z
0.v
)] (20.13)
where 1R
a.v
(m s
1
) is the transfer speed for water vapor. The transfer speed is
larger under unstable than stable conditions.
The resistance to molecular diffusion in the laminar sublayer (s m
1
) is
R
b
= ln
z
0.m
z
0.q
(ScPr)
23
ku
(20.14)
where z
0.m
is the surface roughness length for momentum, dened in Section 8.3.
For vegetated surfaces, z
0.m
z
0.q
is approximately 100. For snow, ice, water, and
bare soil, z
0.m
z
0.q
ranges from-1 to 3. For rough surfaces, z
0.m
z
0.q
can be as large
as 10
5
(Garratt and Hicks 1973; Brutsaert 1991; Ganzeveld and Lelieveld 1995).
The Schmidt number for particles is Sc
pi
=
a
D
pi
and for gases is Sc
q
=
a
D
q
.
667
The Prandtl number is Pr =
a
c
p.m
a
. These three terms were dened in (15.36),
(16.25), and (16.32), respectively.
The resistance of gases to surface interactions is affected by chemical and bio-
logical processes, which vary with surface composition. A parameterization for
gas deposition to vegetation is described here (Wesely 1989; Walmsley and Wesely
1996). In this treatment, the total surface resistance (s m
1
) is approximated with
R
s
=
1
R
stom
+ R
meso
+
1
R
cut
+
1
R
conv
+ R
surf
+
1
R
canp
+ R
soil
1
(20.15)
where R
stom
is the leaf stomata resistance, R
meso
is the leaf mesophyll resistance,
R
cut
is the leaf cuticle resistance, R
conv
is the resistance due to buoyant convection in
canopies, R
surf
is the resistance due to leaves, twigs, bark, or other exposed surfaces
in the lower canopy, R
canp
is the resistance due to canopy height and density, and
R
soil
is the resistance due to soil and leaf litter at the ground surface.
The stomata, mesophyll, and cuticle resistances account for deposition of gases
into openings on leaf surfaces, into liquid water within leaves, and onto outer leaf
surfaces, respectively. A plant cuticle is a waxy outer covering of a plant surface,
formed fromcutin. The cuticle is relatively impermeable to water vapor, preventing
signicant water loss fromthe leaf. Water vapor, carbon dioxide, oxygen, and other
gases can pass through the cuticle into or out of plant leaves through stomata.
Stomata are openings, scattered throughout the cuticle, that extend into the leafs
epidermis, or underlying skin. Stomata are important for allowing carbon dioxide
to enter the middle leaf, called the mesophyll. The mesophyll contains empty spaces
and chloroplasts. Chloroplasts are cells in the leaf in which photosynthesis takes
place. During photosynthesis, carbon dioxide and water are converted to molecular
oxygen and organic material. The oxygen escapes the mesophyll to the atmosphere
through the stomata. Liquid water, brought through xylem from the plants roots
to the mesophyll, evaporates from the surfaces of chloroplasts into the open space
of the mesophyll. This vapor diffuses, often through the stomata, to the open
atmosphere. Gases from the atmosphere can also diffuse through the stomata to
the mesophyll, where they may deposit onto chloroplast surfaces.
The leaf stomata resistance (s m
1
) accounts for the loss of a gas due to deposition
within stomata. The actual stomata resistance depends on the plant species, but a
parameterization that represents an average resistance for many species is
R
stom.q
= R
min
1 +
200
S
f
+0.1
400
T
a.c
(40 T
a.c
)
D
v
D
q
(20.16)
(Baldocchi et al. 1987; Wesely 1989), where R
min
is the minimum bulk canopy
stomata resistance for water vapor, S
f
is the surface solar irradiance (W m
2
), T
a.c
is the surface air temperature (
C), and D
v
D
q
is the ratio of the diffusion coefcient
of water vapor to that of gas q. Minimal resistances for water vapor are given in
Appendix Table B.11 for different land-use types and seasons. Diffusion-coefcient
ratios for different gases are given in Appendix Table B.12. In (20.16), T
a.c
ranges
from 0 to 40

C. Outside this range, R
stom.q
is assumed to be innite. During the
668
20.2 Dry deposition
Table 20.3 Cuticle resistances R
cut
(s m
1
) for SO
2
, O
3
, and any gas q
R
cut,SO
2
Nonurban Urban R
cut,O
3
R
cut,q
Dry R
dry
cut.SO
2
R
dry
cut.SO
2
R
dry
cut.O
3
R
dry
cut.q
Dew 100 50
1
3000
+
1
3R
cut.0
1

1
3R
dry
cut.q
+
H
q
10
7
+
f
0.q
R
dew
cut.O
3
1
Rain
1
5000
+
1
3R
cut.0
1
50
1
1000
+
1
3R
cut.0
1

1
3R
dry
cut.q
+
H
q
10
7
+
f
0.q
R
dew
cut.O
3
1
Urban refers to areas with little vegetation, such as urban areas. The cuticle resistance for a dry
surface is given in (20.18).
Sources: Wesely (1989), Walmsley and Wesely (1996).
night, stomata close, and resistance to deposition increases to innity. The equation
simulates this effect, predicting that R
stom.q
approaches 10
6
when S
f
= 0.
Leaf mesophyll resistance depends on the ability of a gas to dissolve and react
in liquid water on the surface of plant cells within the mesophyll of a leaf. The
ability of a gas to dissolve depends on its effective Henrys constant H
q
and on a
reactivity factor for gases, f
0.q
. The effective Henrys lawconstant is the ratio of the
molarity of a dissolved substance plus its dissociated ions to the gas-phase partial
pressure of the substance. It differs from a regular Henrys law constant (Section
17.4.1), which is the ratio of the molarity of the dissolved substance alone to its
partial pressure. The reactivity factor accounts for the ability of the dissolved gas
to oxidize biological substances once in solution. A value of f
0.q
= 1 indicates the
dissolved gas is highly reactive, and f
0.q
= 0 indicates it is nonreactive. The leaf
mesophyll resistance is estimated as
R
meso.q
=
q
3000
+100 f
0.q
1
(20.17)
where values of H
q
and f
0.q
are given in Appendix Table B.12 for several gases.
The leaf cuticle resistance accounts for the deposition of gases to leaf cuticles
when vegetation is healthy, and to other vegetated surfaces in the canopy when
it is not. Table 20.3 gives expressions of this resistance for sulfur dioxide, ozone,
and any other gas when the leaf cuticle is dry, moist from dew, and wet from
rain.
The leaf cuticle resistance over a sufciently dry surface (s m
1
), required in
Table 20.3, is approximated as
R
dry
cut.q
= R
cut.0
q
10
5
+ f
0.q
1
(20.18)
where R
cut.0
is given in Appendix Table B.11 for several seasons and land-use
categories.
669
The resistance of gases to buoyant convection (s m
1
) is estimated for all gases
as
R
conv
= 100
1 +
1000
S
f
+10
1
1 +1000s
t
(20.19)
where s
t
is the slope of the local terrain, in radians. The resistance due to deposition
on leaves, twigs, bark, and other exposed surfaces (s m
1
) is estimated for any gas
q as
R
surf.q
=
10
5
H
q
R
surf.SO
2
+
f
0.q
R
surf.O
3
1
(20.20)
where R
surf.SO
2
and R
surf.O
3
are given in Appendix Table B.11.
The in-canopy resistance (s m
1
) is a function of canopy height (h
c
) and one-sided
leaf area index (L
T
), which was dened in Section 8.3 as the area of leaf surface
between the ground and canopy per unit area of underlying ground. Appendix
Table B.11 estimates values of in-canopy resistance. Another parameterization of
this resistance is
R
canp
=
b
c
h
c
L
T
u
(20.21)
where b
c
= 14 m
1
, obtained from data (Erisman et al. 1994).
Finally, the resistance due to soil and leaf litter at the ground surface (s m
1
) is
estimated with
R
soil.q
=
10
5
H
q
R
soil.SO
2
+
f
0.q
R
soil.O
3
1
(20.22)
where R
soil.SO
2
and R
soil.O
3
are given in Appendix Table B.11.
20.3 DRY DEPOSITION AND SEDIMENTATION
CALCULATIONS
Figure 20.3 shows particle sedimentation (fall) speed, particle dry-deposition speed
in the absence of sedimentation, and particle total dry-deposition speed as a func-
tion of diameter. For small particles, the nonsedimentation component of the dry-
deposition speed exceeds the sedimentation component, because particle diffusion
affects small particles more than does gravity. For large particles, sedimentation
dominates, because the effect of gravity exceeds that of particle diffusion. The total
dry-deposition speed is its smallest for particles of about 1 m in diameter; nei-
ther diffusion nor gravity dominates for particles near that size. The sedimentation
speed levels off at large diameters due to the physical-property correction term in
the sedimentation equation. At small particle diameters, the slope of the sedimen-
tation speed also decreases due to the correction for particle resistance to motion,
which accounts for the fact that particles can move more freely through air as their
size decreases.
670
20.3 Dry deposition and sedimentation calculations
10
6
10
4
10
2
10
0
10
2
10
4
0.01 0.1 1 10 100 1000
S
p
e
e
d

(
c
m

s
1

)
Dry-deposition
speed without
sedimentation
Sedimentation speed
Total dry-deposition
speed
Figure 20.3 Particle total dry-deposition speed and
its components versus particle diameter. Conditions
were z
r
= 10 m, z
0,m
= 0.0001 m, T(z
r
) = 290.1 K,
T(z
0,m
) = 290 K, p
a
(z
r
) = 999 hPa, |v
h
(z
r
)| =
10 m s
1
, and
p
= 1.0 g cm
3
. The near-ground
temperature was taken at z
0,m
instead of z
0,h
. Sur-
face resistance was ignored.
10
11
10
7
10
3
10
1
10
5
0.01 0.1 1 10 100 1000
Knudsen
number
Total dry-deposition speed (cm s
1
)
Reynolds
number
Diffusion coefficient (cm
2
s
1
)
Figure 20.4 Particle total dry-deposition speed
from (20.11), Reynolds number, Knudsen number
for air, and particle diffusion coefcient versus par-
ticle diameter. Conditions were z
r
= 10 m, z
0,m
=
0.0001 m, T(z
r
) =292 K, T(z
0,m
) =290 K, p
a
(z
r
) =
999 hPa, |v
h
(z
r
)|= 10 m s
1
, and
p
= 1.0 g cm
3
.
The near-ground temperature was taken at z
0,m
instead of z
0,h
.
Figure 20.4 shows the calculated particle diffusion coefcient, Knudsen number
for air, Reynolds number, and particle total dry-deposition speed as a function of
size, for the conditions described in the gure. The diffusion coefcient decreased
with increasing particle size, as expected, since the diffusion coefcient is inversely
proportional to the radius of the particle, as shown in (15.29). The Knudsen number
also decreased with increasing size, since it is inversely proportional to the radius,
as shown in (15.23). The Reynolds number increased with increasing size because it
is proportional to the radius and fall speed, both of which increased with increasing
671
10
3
10
2
10
1
10
0
10
1
0 10 20 30 40 50
10 g mol
1
, 10 m s
1

130 g mol
1
, 10 m s
1
10 g mol
1
, 0 m s
1
130 g mol
1
, 0 m s
1
D
r
y

d
e
p
o
s
i
t
i
o
n

s
p
e
e
d

(
c
m

s
1
)
Surface resistance (s cm
1
)
10
2
10
1
10
0
10
1
0 10 20 30 40 50
10 g mol
1
, 10 m s
1
130 g mol
1
, 10 m s
1
10 g mol
1
, 0 m s
1
130 g mol
1
, 0 m s
1
D
r
y

d
e
p
o
s
i
t
i
o
n

s
p
e
e
d

(
c
m

s
1
)
Surface resistance (s cm
1
)
(a) (b)
Figure 20.5 Gas dry-deposition speed versus surface resistance when (a) z
0,m
= 3 m and
(b) z
0,m
= 0.01 m, and when the molecular weight and wind speed varied, as indicated in
the gure legend. Also, z
r
= 10 m, T(z
r
) = 288 K, T(z
0,m
) = 290 K, and p
a
(z
r
) = 999 hPa.
size. The dry-deposition speed decreased then increased with increasing size, as in
Fig. 20.3.
Figures 20.5 (a) and (b) show gas deposition speeds versus surface resistance
and as a function of molecular weight, wind speed, and surface roughness length
for momentum in an unstable atmosphere. The curves show that surface resistance
affected dry-deposition speed most when the surface roughness length was large
and the wind speed was high. Molecular weight had relatively little effect on dry-
deposition speed, except when the wind speed was zero. Surface roughness length
affected dry-deposition speed primarily when the surface resistance was low. Wind
speed affected dry-deposition speed at both roughness lengths.
20.4 AIRSEA FLUX OF CARBON DIOXIDE AND
OTHER GASES
The ux of carbon dioxide between the atmosphere and ocean is of particular
importance on a global scale because atmospheric CO
2
has been increasing due
to anthropogenic emission, and dissolution into the ocean is a major mechanism
of its removal. As discussed in Chapter 17, Henrys law states that, over a dilute
solution, the partial pressure exerted by a gas is proportional to the molality of the
dissolved gas in solution. Thus, the more CO
2
in the atmosphere, the more that it
dissolves in the ocean. The quantity of dissolved gas, though, is also a function of
pH, temperature, and the concentration of other ions, such as calcium, in ocean
water. Further, since the atmosphere and ocean are not in perfect equilibrium,
the transfer of CO
2
to (from) the ocean following its increase (decrease) in the
atmosphere is a time-dependent process that depends on the dry-deposition speed.
The uxes of gases aside from CO
2
between the atmosphere and ocean are also
important since the ocean serves as a sink for all gases and a source for some. An
important gas emitted from the ocean, for example, is dimethylsulde (DMS) (e.g.,
Kettle and Andreae 2000), and deposition to the ocean surface is one of the main
removal mechanisms of, for example, ozone (e.g., Galbally and Roy 1980; Chang
et al. 2004).
672
20.4 Airsea flux of carbon dioxide and other gases
20.4.1 Numerical solution to airsea exchange
In this section, a method of solving equations for the exchange of gases between
the ocean and atmosphere is derived. The change in concentration of any gas q
between the atmosphere and ocean due to surface dissolution and evaporation can
be described by
C
q.t
= C
q.th
+
hV
d.gas.q
Lz
a
c
q.T.t
H
q
C
q.t
(20.23)
where the subscripts t and t h indicate the current time and one time step back-
ward, respectively, h is the time step (s), C
q
is the atmospheric mole concentration
of the gas (mol cm
3
-air), c
q.T
is the mole concentration of the dissolved gas plus its
dissociation products in seawater (mol cm
3
-air = mol cm
3
-seawater since sea-
water occupies the same volume as the air it displaces), V
d.gas.q
is the dry deposition
speed of the gas (cm s
1
), Lz
a
is the thickness of the atmospheric layer through
which C
q
is averaged (cm), and H
q
is a dimensionless effective Henrys law con-
stant (mol mol
1
). Equation (20.23) is written implicitly; thus, c
q.T.t
and C
q.t
are
unknown. If the terms were written explicitly (with time subscript t h) on the
right-most side of the equation, the nal concentration C
q.t
could become negative
at a long time step or high deposition speed.
The mole concentration of a gas is related to its partial pressure by
C
q
=
p
q
R
T
(20.24)
where p
q
is the partial pressure of the gas (atm), R

(82.058 cm
3
atm mol
1
K
1
), and T is absolute air temperature (K). The mole
concentration of a gas dissolved in seawater (mol cm
3
) is related to its molality
in seawater by
c
q.T
=
dw
m
q.T
(20.25)
where m
q.T
is the molality (mol-solute per kilogramof dilute water) of the dissolved
gas plus its dissociation products, if any, and
dw
is the density of dilute water, which
is close to 0.001 kg cm
3
, but varies slightly with temperature. Dilute water instead
of seawater is used in this equation since molality is dened as the moles of solute
per kilogram of solvent (water), not of solution (seawater). If the gas dissolves and
dissociates twice in solution, as carbon dioxide does by the following reactions,
CO
2
(g) +H
2
O(aq)
H
H
2
CO
3
(aq)
K
1
H
+
+HCO

3
K
2
2H
+
+CO
2
3
(20.26)
then the dimensionless effective Henrys coefcient of the gas in seawater is
H
q
=
dw
R
TH
s
q
1 +
K
s
1.q
m
H
+
1 +
K
s
2.q
m
H
+
(20.27)
673
where H
s
q
is the Henrys law coefcient of the gas measured in seawater but written
in terms of dilute water (mol kg
1
-dw atm
1
), K
s
1.q
and K
s
2.q
are the rst and
second dissociation coefcients, respectively, of the dissolved gas (mol kg
1
-dw),
and m
H
+ is the molality of free (dissociated) hydrogen ion in dilute water (mol
kg
1
-dw). The unit mol kg
1
-dw for a rate coefcient indicates that its units
have been converted, if necessary, from moles per kilogram of seawater to moles
per kilogram of dilute water. In the case of carbon dioxide dissolving in seawater,
expressions for the Henrys law and rst and second dissociation coefcients are
ln H
s
CO
2
mol
kg-dw atm
60.2409 +93.4517
100
T
w
+23.3585 ln
T
w
100
+S
0.023517 0.023656
T
w
100
+0.0047036
T
w
100
(20.28)
ln K
s
1.CO
2
mol
kg-dw
=

sw
dw
m
H
+
m
H
+
.SWS
2.18867
2275.0360
T
w
1.468591 ln T
w
+
0.138681
9.33291
T
w
S
0.5
+0.0726483S
0.00574938S
1.5
(20.29)
ln K
s
2.CO
2
mol
kg-dw
=

sw
dw
m
H
+
m
H
+
.SWS
0.84226
3741.1288
T
w
1.437139 ln T
w
+
0.128417
24.41239
T
w
S
0.5
+0.1195308S
0.00912840S
1.5
(20.30)
respectively (Millero 1995 and references therein). In these equations, T
w
is the
absolute temperature (K) of water, S is seawater salinity (parts per thousand by
mass),
sw
is the density of seawater (kg cm
3
), and m
H
+
.SWS
is the molality of
total free plus associated hydrogen ion in solution based on the seawater scale
(mol kg
1
-dw).
The salinity and density of seawater are
S = 1000
c
q
m
q
c
H
2
O
m
H
2
O
c
q
m
q
(20.31)
sw
=

dw
1 0.001S
(20.32)
respectively, where the summation in the salinity equation is over all components q
in seawater, including dilute water, c is mole concentration in seawater (mol cm
3
-
water), m is molecular weight (g mol
1
), and
dw
is the density of dilute water
674
(kg cm
3
). A typical salinity of seawater is 35 ppth, giving a typical seawater
density of 0.001036 kg cm
3
. The ratio
sw
dw
= 1(1 0.001S) is necessary to
convert measured reaction rate coefcients from mol kg
1
-sw to mol kg
1
-dw.
The molality (mol kg
1
-dw) of free hydrogen in seawater is related to the pH of
seawater by
m
H
+ =
10
pH
dw
1000
(20.33)
where 1000 is the number of cubic centimeters per liter. The pH of seawater
generally ranges from 7.8 to 8.3. The parameter m
H
+
.SWS
is approximately the
sum of the molalities of the free hydrogen ion, the bisulfate ion, and dissolved
hydrouoric acid in seawater (Millero 1995). At T
w
= 298.15 K and S = 35 ppth,
the ratio m
H
+m
H
+
.SWS
is approximately 0.763. This ratio is needed in the rate
equations since the equations were derived for seawater solutions.
The dry deposition speed in (20.23) is the same as that in (20.10), except that
the surface resistance over the ocean is
R
s.q
=
1
r.q
H
q
k
w.q
(20.34)
where
r.q
is the dimensionless enhancement of gas transfer to seawater due to
chemical reaction on the ocean surface, and k
w.q
is the transfer speed of a chemically
unreactive gas through a thin lm of water at the ocean surface to the mixed layer
of the ocean (cm s
1
). For extremely soluble gases, such as HCl, H
2
SO
4
, HNO
3
,
and NH
3
, the dimensionless effective Henrys constant H
q
is large and
r.q
may
be large, so the surface resistance (R
s.q
) is small, and the dry deposition speed
is limited only by the aerodynamic resistance (R
a.q
) and resistance to molecular
diffusion (R
b.q
). On the other hand, for slightly soluble gases, such as CO
2
, CH
4
,
O
2
, N
2
, and N
2
O, H
q
is relatively small,
r.q
1, so R
s.q
is large and the dry
deposition speed is controlled primarily by k
w.q
.
The transfer speed through a thin lm of water on the ocean surface is affected
by the gass dissolution in and molecular diffusion through the lm and surfactants
and bursting of bubbles on the surface of the lm. Although the transfer speed
depends on several physical processes, parameterizations of k
w.q
to date have been
derived only in terms of wind speed. One such parameterization is
k
w.q
=
1
3600
0.17|v
h.10
|S
23
rw.q
|v
h.10
| 3.6ms
1
0.612S
23
rw.q
+(2.85|v
h.10
| 10.262)S
12
rw.q
3.6ms
1
- |v
h.10
| 13 ms
1
0.612S
23
rw.q
+(5.9|v
h.10
| 49.912)S
12
rw.q
13 ms
1
- |v
h.10
|
(20.35)
(Liss and Merlivat 1986), where k
w.q
is in units of cm s
1
, the constant 3600
converts cmhr
1
to cms
1
, |v
h.10
| is the wind speed at 10 mabove sea level (ms
1
),
and
S
rw.q
=
Sc
w.CO
2
.20
C
Sc
w.q.T
s.c
(20.36)
675
is the ratio of the Schmidt number in water of carbon dioxide at 20
C to that of
gas q at the current temperature. An equation for the Schmidt number in air was
given in (16.25). A second parameterization for the transfer speed of a gas through
water is
k
w.q
=
0.31|v
h.10
|
2
S
12
rw.q
3600
(20.37)
(Wanninkhof 1992), where k
w.q
is in units of cm s
1
. An expression for the dimen-
sionless Schmidt number of CO
2
in seawater, valid from 0 to 30

C, is
Sc
w.CO
2
= 2073.1 147.12T
s.c
+3.6276T
2
s.c
0.043219T
3
s.c
(20.38)
(Wanninkhof 1992), where T
s,c
is the temperature of seawater in degrees Celsius.
Equation (20.23) gives the time-dependent change in gas mole concentration due
to transfer to and from ocean water. Here, a method of solving (20.23) is shown.
Rearranging (20.23) to solve for gas concentration give
C
q.t
=
C
q.th
+
hV
d.gas.q
Lz
a
c
q.T.t
H
q
1 +
hV
d.gas.q
Lz
a
(20.39)
where the mole concentration in seawater, c
q.T.t
, is still unknown. Mole balance
between the ocean and atmosphere requires
c
q.T.t
D
l
+C
q.t
Lz
a
= c
q.T.th
D
l
+C
q.th
Lz
a
(20.40)
where D
l
is the depth (cm) of the ocean mixed layer. Substituting (20.39) into
(20.40) and solving give
c
q.T.t
=
c
q.T.th
+
hV
d.gas.q
C
q.th
D
l
1 +
hV
d.gas.q
Lz
a
1 +
hV
d.gas.q
D
l
H
1 +
hV
d.gas.q
Lz
a
(20.41)
(Jacobson 2005c). Substituting (20.41) into (20.39) gives the nal gas concentra-
tion for the time step. Equations (20.39) and (20.41) are referred to as the Ocean
Predictor of Dissolution (OPD) scheme. These equations conserve moles exactly,
are noniterative, and cannot produce a negative gas or water concentration, regard-
less of the time step.
Once dissolved in ocean water, a chemical diffuses and chemically reacts. Many
seawater reactions have been identied (e.g., Stumm and Morgan 1981; Butler
1982; Millero 1995). Here, just one is described. When carbon dioxide dissociates
676
in water by (20.26), its dissociation products may combine with cations and crys-
tallize, removing the ions from solution. The most common carbonate minerals
in the ocean are calcite and aragonite, both of which have the chemical formula
CaCO
3
(s). The reversible reaction forming calcite or aragonite is
Ca
2+
+CO
2
3

CaCO
3
(s) (20.42)
Calcite or aragonite forms only when
m
Ca
2+m
CO
2
3
> K
s
CaCO
3
(T) (20.43)
where
m
CO
2
3
=
c
CO
2
.T
dw
K
s
1.CO
2
K
s
2.CO
2
m
H
+m
H
+ + K
s
1.CO
2
m
H
+ + K
s
1.CO
2
K
s
2.CO
2
(20.44)
is the molality (mol kg
1
-dw) of the carbonate ion in seawater, m
Ca
2+ 0.01028
mol kg
1
-dw is the molality of the calcium ion in seawater (derived from Table
14.1), and K
s
CaCO
3
(T) is the solubility product (equilibrium coefcient) of the reac-
tion in seawater. The solubility product of calcite at 25
C and 35 parts per thou-

sand salinity in seawater (with the solubility product converted to dilute water) is
K
s
CaCO
3
(T) =4.427 10
7
mol
2
kg
2
-dw. That for aragonite is 6.717 10
7
mol
2
kg
2
-dw(derived fromMillero 1995). Thus, the formation of calcite is favored over
the formation of aragonite when only solubility products are considered.
Example 20.1
If the molality of total dissolved carbon is 0.0023 mol kg
1
-dw, pH = 8.0,
T = 298.15 K, and
dw
= 1 g cm
3
, estimate the molality of aragonite formed
in seawater assuming equilibrium.
SOLUTION
From (20.29), K
1.CO
2
= 1.125 10
6
mol kg
1
-dw; from (20.30), K
2.CO
2
=
9.603 10
10
mol kg
1
-dw; from (20.33), m
H
+ = 1 10
8
mol kg
1
; and
from (20.44), m
CO
2
3
= 1.999 10
4
mol kg
1
-dw. Taking the difference
m
Ca
2+
m
CO
2
3
K
eq
(T) with m
Ca
2+ = 0.01028 mol kg
1
-dw gives 1.38 10
6
mol kg
1
-dw of aragonite in seawater.
The product of the measured molalities of the calcium ion and the carbonate
ion in seawater exceed the solubility product of calcite near the ocean surface (e.g.,
Stumm and Morgan 1981; Butler 1982). Part of the reason is that magnesium
poisons the surface of a growing calcite crystal, creating a magnesium-calcite crystal
that is more soluble than calcite alone, increasing the dissolution rate of calcite.
For this reason, aragonite is often found in higher concentration in the mixed layer
than is calcite.
677
280
300
320
340
360
380
8.1
8.15
8.2
8.25
8.3
1750 1800 1850 1900 1950 2000
C
O
2

m
i
x
i
n
g

r
a
t
i
o

(
p
p
m
v
)
S
u
r
f
a
c
e

o
c
e
a
n

p
H
Year
Figure 20.6 Modeled (-) versus measured (- - - -) CO
2
mix-
ing ratio and modeled surface ocean pH between 1750
and 2004 in a one-dimensional atmosphereocean calcula-
tion. The model treated 38 ocean layers of 100-m thickness
and one atmospheric layer divided into two compartments:
landair and oceanair. CO
2
was emitted in the land compart-
ments, partitioned each time step between the landair and
oceanair compartments, and transferred between the ocean
air compartment and surface ocean. Dissolved CO
2
diffused
vertically in the ocean. Ocean chemistry was calculated in all
layers. The CO
2
mixing ratio data were from Friedli et al.
(1986) up to 1953 and from Keeling and Whorf (2003) for
19582003. From Jacobson (2005c).
20.4.2 Simulation of ocean acidication
Figure 20.6 shows results from a one-dimensional model simulation of historic
and future ocean composition and pH in which the airsea exchange algorithm
just described was used. For the simulation, the ocean consisted of 38 100-m-
thick layers (extending to 3800 m depth). The atmosphere, which consisted of
one layer, was divided into a landair and an oceanair compartment. CO
2
was
emitted into the landair compartment. The added CO
2
was then instantaneously
mixed between the oceanair and landair compartments in a mole-conserving
manner assuming the ocean comprised 71.3 percent of the Earths surface. Airsea
exchange equations were then solved with a time step of one day, changing ocean
and air carbon concentrations and ocean pH. Each time step, vertical diffusion and
equilibrium chemistry were solved in the ocean for dozens of chemicals.
Historic CO
2
emission from 1751 to near present was obtained from Marland
et al. (2003). CO
2
was solved for prognostically; it was not nudged or assimilated.
If the historic CO
2
results are correct, Fig. 20.6 suggests that ocean pH may have
decreased from about 8.247 in 1751 to 8.136 in 2004, for a 26 percent increase
in the hydrogen ion content of the ocean. A future simulation was also run using
Special Report on Emission Scenarios (SRES), A1B CO
2
future emission factors
(Nakicenovic et al. 2000). Based on the result, the hydrogen ion concentration in
2100 (pH = 7.846) may be a factor of 2.5 greater than in 1751 (pH = 8.247).
678
These results indicate that carbon dioxide buildup in the atmosphere leads to ocean
acidication, a general nding consistent with that from other studies (e.g., Brewer
1997; Caldeira and Wickett 2003). The simulations also uncovered the possibility
that ocean acidication due to CO
2
may slightly increase the atmospheric concen-
trations of non-CO
2
atmospheric acids, such as hydrochloric acid, nitric acid, and
sulfurous acids, but may decrease the concentration of the base ammonia.
20.5 SUMMARY
In this chapter, particle sedimentation and gas and particle dry deposition were
discussed. Sedimentation is affected by gravity and air viscosity. At very small and
at very large particle diameters, sedimentation speeds are modied by particle resis-
tance to motion and physical-property effects, respectively. Particle sedimentation
speeds decrease with decreasing particle size, but dry-deposition speeds increase
with decreasing size due to enhanced molecular diffusion at small size. Particle
dry-deposition speeds are inuenced by meteorological variables and diffusion at
small diameters and by sedimentation at large diameters. Gas dry-deposition speeds
are inuenced by meteorological variables and surface interactions. Such inter-
actions can be parameterized as a function of leaf-stomata, leaf-mesophyll, leaf-
cuticle, buoyant-convective, exposed-surface, canopy, and soil resistances. Finally,
a method of calculating the ux of gas between the atmosphere and ocean was
discussed. Carbon dioxide uxes change ocean pH and chemical composition and,
in turn, are affected by them.
20.6 PROBLEMS
20.1 Calculate the fall speed of a 10-m-radius soil particle of density 2.5 g cm
3
when T = 230 K and (a) p
a
= 300 hPa and (b) p
a
= 10 hPa. Assume particles
are spherical and the air is dry.
20.2 Briey describe the three resistances affecting gas dry-deposition speeds.
What condition(s) give(s) maximum values for each resistance?
20.3 Howdo increases in surface heating and instability, respectively, affect indi-
vidual components of surface resistance?
20.4 Compare the CO
2
transfer speed at 20

C from (20.33) and (20.35) when the
wind speed is (a) 3 ms
1
and (b) 10 ms
1
. Which parameterization should
result in a perturbation to atmospheric CO
2
reaching equilibrium with the
ocean faster?
20.5 Terrain signicantly affects particle dry-deposition speed. Write a computer
script to calculate dry-deposition speeds of spherical particles 0.01 m in
diameter when z
0,m
= 0.0001 m (ocean) and when z
0,m
= 5 m (urban area).
At z
0,m
= 0.0001 m, assume u
= 0.48 m s
1
and L = +180 m. At z
0,m
= 5 m,
679
assume u
= 0.8 m s
1
and L = +3000 m. Also, assume p
d
= 1000 hPa and
T = 292 K at z
r
= 10 m, and
v
= 1 g cm
3
.
20.6 Submicrometer atmospheric particles require signicant time to fall out of
the atmosphere. Write a computer script to calculate howlong a 0.1-m- and
a 10-m-diameter liquid water particle take to sediment from an altitude of
2 km to 1 km. Assume T = 280 K and p
d
= 850 hPa are constant. Assume
drops are spherical and
v
= 1 g cm
3
.
20.7 Write a computer script to calculate the time required for molecules of ozone
and of carbon monoxide to dry-deposit to the surface froma reference height
of z
r
= 10 m, assuming R
s
= 500 s m
1
for O
3
and R
s
= 3000 s m
1
for CO.
Assume p
d
= 1000 hPa and T = 290 K at the ground surface. Also assume
that, at z
r
= 10 m, p
d
= 999 hPa, T = 288 K, u = 7 m s
1
, and v = 7 m s
1
.
Let z
0,m
= 1 m. Use the temperature, pressure, etc. at 10 m to calculate other
atmospheric variables.
20.8 Write a computer script to compare the dry-deposition speed of a 10-m-
diameter particle with that of a 0.1-m-diameter particle to determine which
is inuenced more by surface resistance. Assume V
f
= 0.0031 m s
1
for the
10-m particle and V
f
= 8.6 10
7
m s
1
for the 0.1-m particle. Assume
the other conditions are the same as in Problem 20.7.
20.9 Meteorological conditions signicantly affect gas dry-deposition speeds
when the resistance to surface interactions (R
s
) is small. Write a computer
script to calculate the difference in the dry-deposition speed of water vapor
when R
s
= 100.0 s m
1
in the following two cases: (a) when u = 15 m s
1
,
v = 15 m s
1
, p
d
= 999 hPa, and T = 292 K at z
r
= 10 m, and (b) when u =
2 m s
1
, v = 2 m s
1
, p
d
= 999 hPa, and T = 288 K at z
r
= 10 m. In both
cases, assume p
d
= 1000 hPa and T = 290 K at the surface and z
0,m
= 0.1 m.
20.10 Write a computer script to calculate surface resistance with equations in
Section 20.2. (a) Calculate the total surface resistance for SO
2
during sum-
mer over a coniferous forest when S
f
= 600 W m
2
, T
a.c
= 15
C, and s
t
= 0.
(b) Assume S
f
varies as a sine function between 6 a.m. and 6 p.m., peaking
with S
f
= 800 W m
2
at noon and equaling zero at night. Plot surface resis-
tance versus time for a 24-h period, assuming the same remaining conditions
in part (a). (c) Repeat part (b) for O
3
. Comment on the diurnal variations of
surface resistances for both gases.
20.11 Write a computer script to solve (20.39) and (20.41). Use the program to
calculate the change in concentration over time of ozone in water and air
assuming T =288 K, a wind speed of 10 ms
1
, a transfer speed from (20.37),
a dry-deposition speed affected only by surface resistance, a Henrys con-
stant from Appendix Table B.7, and a time step of one hour. Assume the
atmosphere depth is 8 km and the ocean depth is 100 m. Assume the ini-
tial mixing ratio in air is 40 ppbv and zero in water. Assume air density
is uniform through the atmospheric layer at 1.2 kg m
3
. Show results over
5 years.
680
21
Model design, application, and testing
I
n previous chapters, physical, chemical, radiative, and dynamical processes and
numerical methods to simulate those processes were described. Here, steps in
model design, application, and testing are discussed. The most important of these
steps are to dene the goals of a modeling study, select appropriate algorithms,
obtain input and emission data, compare model predictions with data, and analyze
results. To illustrate these steps, the design of a model is briey discussed, and
statistical and graphical comparisons of predictions of the model with data are
shown. This chapter integrates numerical methods discussed in previous chapters
with model development and analysis procedures.
21.1 STEPS IN MODEL FORMULATION
Model design, application, and testing require several steps. These include (1) de-
ning and understanding the problem of interest, (2) determining the spatial and
temporal scale of the problem, (3) determining the dimension of the model,
(4) selecting the physical, chemical, and/or dynamical processes to simulate,
(5) selecting variables, (6) selecting a computer architecture, (7) codifying and
implementing algorithms, (8) optimizing the model on a computer architecture,
(9) selecting time steps and intervals (10) setting initial conditions, (11) setting
boundary conditions, (12) obtaining input data, (13) obtaining ambient data for
comparison, (14) interpolating input data and model predictions, (15) developing
statistical and graphical techniques, (16) comparing results with data, (17) running
sensitivity tests and analyzing the results, and (18) improving algorithms. Each of
these steps is discussed below.
21.1.1 Dening the purpose of the model
The rst step in model development is to dene the scientic, regulatory, or
computational problem of interest. Example scientic topics include determining
the effects of (1) gas and particle emission on ambient pollutant concentrations,
(2) gas and particle emission on global and regional climate, (3) gas and particle
emission on stratospheric ozone, (4) gas and particle emission on acid deposi-
tion, (5) global climate change on regional air quality, (6) aerosol particles on local
weather patterns, (7) aerosol particles on photolysis rates of gases, (8) aerosol parti-
cles on cloud cover, (9) aerosol particles on atmospheric stability, (10) soil moisture
681
on atmospheric stability, boundary layer height and pollutant concentrations,
(11) global climate change on regional climate weather, (12) climate change on
severe weather, (13) ocean circulation on global climate, and (15) ocean circula-
tion on atmospheric carbon dioxide.
Some topics of interest to regulators and scientists include estimating the effects
of nitrogen oxide and hydrocarbon emission controls on human exposure, chlo-
rouorocarbon emission regulations on stratospheric ozone, and carbon dioxide
and aerosol particle emission controls on global warming. Computational topics
of interest include determining the most efcient computer architecture, in terms
of speed and memory, to run a model on, and developing numerical techniques to
speed algorithms on existing computer architectures.
21.1.2 Determining scales of interest
The second step in model development is to determine the spatial and temporal
scale of interest. Spatial scales include the molecular, micro-, meso-, synoptic, and
global scales (Table 1.1). Urban air pollution, thunderstorms, and tornados are
simulated over micro- to mesoscale domains. Acid deposition, regional climate
change, and hurricanes are simulated over meso- to synoptic-scale domains. Global
climate change is simulated over a global domain. With respect to time, urban air
pollution and weather events are simulated over hours to weeks, acid deposition
events are simulated over days to months, and regional and global climate change
events are simulated over years to centuries.
21.1.3 Determining the dimension of the model
The third step in model development is to determine whether a zero-, one-, two-,
or three-dimensional model is required and whether to nest the model. Three-
dimensional models are ideal, but because such models require enormous computer
time and memory, zero-, one-, and two-dimensional models are often used instead.
A zero-dimensional (0-D) model is a box model in which chemical and/or phys-
ical transformations occur. Gases and particles may enter or leave the box from
any side. Since all material in the box is assumed to mix instantaneously, the con-
centration of each gas and particle is uniform throughout the box. A standard
box model is xed in space. A parcel model is a box model that moves through
space along the direction of the wind. Emission enters the box at different points
along the trajectory. Because a parcel model moves in a Lagrangian sense, it is also
called a Lagrangian trajectory model. Box models have been used to simulate pho-
tochemical reactions that occur in smog chambers, fog production in a controlled
environment, and chemical and physical interactions between aerosol particles and
gases. Parcel models have been used to trace changes in an air parcel as it travels
from ocean to land and through the polar vortex.
A one-dimensional (1-D) model is a set of adjacent box models, stacked verti-
cally or horizontally. Vertical 1-D models may be used to study radiative transfer
682
21.1 Steps in model formulation
with photochemistry, gas and aerosol vertical transport, aerosol optical proper-
ties, aerosol sedimentation, or cloud convection. The main use of a horizontal 1-D
model is to test the 1-D advection diffusion equation. The disadvantage of a 1-D
model compared with a 2-D or 3-D model is that velocities in a 1-D model can-
not be predicted but must be estimated crudely. In addition, a 1-D model either
ignores gas, particle, and potential temperature uxes through lateral (horizontal)
boundaries or roughly parameterizes them. Figure 11.11 shows results from a 1-D
photochemical-radiative model.
A two-dimensional (2-D) model is a set of 1-D models connected side by side.
2-D models can lie in the xy, xz, or yz planes. Advantages of a 2-D over a
1-D model are that transport can be treated more realistically, and a larger spatial
region can be simulated in a 2-D model. 2-D models have been used, for exam-
ple, to simulate dynamics, transport, gas chemistry, and aerosol evolution (e.g.,
Garcia et al. 1992; Rahmes et al. 1998). A global 2-D model may stretch from
the South to the North Pole and vertically. Southnorth and vertical winds in such
a model are predicted or estimated from observations at each latitude and verti-
cal layer. Zonally averaged winds (westeast winds, averaged over all longitudes
for a given latitude and vertical layer) are needed in a 2-D global model. Such
winds are found prognostically or diagnostically. Prognostic winds are obtained,
for example, by writing an equation of motion for the average westeast scalar
velocity at each latitude and altitude. Diagnostic winds are obtained, for example,
by writing prognostic equations for southnorth and vertical scalar velocities, then
extracting the average westeast scalar velocity from the continuity equation for
air.
A three-dimensional (3-D) model is a set of horizontal 2-D models layered on
top of one another. The advantage of a 3-D over a 2-D model is that dynamics and
transport can be treated more realistically in a 3-D model. The disadvantage of a
3-D model is that it requires signicantly more computer time and memory than
does a 2-D model. Nonetheless, studies of urban air pollution are readily carried
out in 3-D, since simulation periods are generally only a few days. Computer-time
limitations are most apparent for global simulations that last months to years.
Because 3-D models represent dynamical and transport processes better than do
0-, 1-, and 2-D models, 3-D models should be used when computer time require-
ments are not a hindrance.
Four main types of 3-D models include global, hemispheric, limited-area, and
nested models. Aglobal model is a model that extends globally in the horizontal but
is capped at its top and bottom by boundaries. A hemispheric model extends glob-
ally in one hemisphere in the westeast direction but has southern and/or northern
boundaries and vertical boundaries. A limited-area model is a nonglobal model
with westeast, southnorth, and vertical boundaries. A nested model is one that
contains two or more 3-D grid domains within it (where a grid domain is a 3-D
model, itself) within it, where all except one of the grid domains (the coarsest-
resolved) receives outside boundary information from a domain with coarser
resolution.
683
Global
R1
R2
R3
Urban
R1
90
60
30
0
30
60
90
180 120 60 0 60 120 180
L
a
t
i
t
u
d
e

(
d
e
g
r
e
e
s
)
Longitude (degrees)
Figure 21.1 An example of nested domains. The largest domain,
Global, covers the Earth. Domain R1 is nested inside the global
domain and receives boundary data from it. Domain R2 is nested
inside domain R1 and receives boundary data fromdomain R1. Three
R3 domains are nested inside domain R2, and each receives bound-
ary data from domain R2. Within each R3 domain are multiple urban
domains that receive boundary information fromthe respective parent
R3 domain.
Grid domains that provide boundary conditions to ner-resolved domains in a
nested model are called parent domains. Domains that receive boundary conditions
from coarser (parent) domains are progeny domains. Since several domains can
be nested within each other (e.g., the Urban, R3, R2, R1, and global domains,
respectively, in Fig. 21.1), a domain can serve as both a parent and a progeny
domain. Each parent domain can also contain multiple progeny domains that are
independent of one another (e.g., the multiple R3 domains within the R2 domain in
Fig. 21.1).
Ideally, the coarsest-resolved domain in a nested model is the global domain,
which does not require boundary information, except at its base and top. In some
nested models, the coarsest domain is a limited-area domain, which needs lat-
eral boundary information. Such information can be obtained from interpolated
meteorological data or from stored global-model results. Variables passed at the
boundaries in a nested model are discussed in Section 21.1.11.
Nesting can be one-way or two-way. Aone-way nested model is a model in which
information is passed only from a parent to a progeny domain. A two-way nested
model is one in which information is passed in both directions. Most nested models
nest one way. One problem with two-way nesting is that it results in perturbations
to velocity, potential temperature, and other variables in the middle of the parent
domain that degrade conservation properties (e.g., conservation of mass, energy,
enstrophy, potential enstrophy, etc.) in the parent domains.
684
21.1.4 Selecting processes
The next step in model development is to select the physical, chemical, and dynam-
ical equations for the model and the best available tools to solve the equations. An
ideal model includes every conceivable process, each simulated with the most accu-
rate solver. Because computer speed and memory are limited, either the number of
processes simulated or the accuracy of individual solutions must be limited.
Seven major groups of processes simulated in atmospheric models are meteo-
rological, transport, cloud, radiative, gas, aerosol, and surface processes. When a
model is developed, it is necessary to decide whether one or more of these groups
can be excluded from the simulation or replaced by measurements.
Many air pollution models interpolate meteorological elds from observations,
ignore the effects of clouds, and/or ignore the effects of aerosol particles. The
advantage of using interpolated meteorological elds is that, if sufcient data are
available, an interpolated eld is more accurate than is a prognostic eld. The dis-
advantage is that measured meteorological data are usually available at only a few
locations, most of which are near the surface. Data aloft are scarce or nonexistent.
Also, a model that uses a pre-existing database of meteorology cannot be used to
predict (forecast) weather or air pollution in the future. It can examine only events
in the past (hindcast).
Finally, many models ignore the effects of hydrometeor particles and/or aerosol
particles. In some urban locations, clouds do not form frequently during the
summer; so, cloud formation is neglected. Aerosol particles often, although not
always, have a minor impact on ozone formation, so studies of ozone often neglect
aerosol particles. Modeling studies of clouds often neglect the effect of aerosol par-
ticles on cloud formation. This assumption may lead to errors since clouds form
physically on aerosol particles.
21.1.5 Selecting variables
Once processes are selected, variables must be chosen. If a model treats meteo-
rology, some of the variables solved for include air temperature, pressure, density,
velocity, and geopotential, and the concentrations of water-vapor, liquid water,
and ice. Meteorological models also calculate energy uxes, moisture uxes, and
temperature at soil, vegetation, and ocean surfaces. If a model includes trace gases,
it calculates concentrations of each gas. If the model includes aerosol particles, it
may calculate particle number concentration and the concentration of each individ-
ual component in each particle size bin. If a model includes radiative processes, it
often solves for heating rates. Many intermediate variables are also stored, but not
permanently. Photolysis rate coefcients, extinction coefcients, particle growth
rates, gas dry-deposition speeds, entrainment rates, and pressure-gradient forces
are variables that are stored temporarily.
When a 3-D model is run at high spatial resolution and/or includes many vari-
ables, computer memory limits must be considered. For example, an atmospheric
model that treats meteorology, gases, size-resolved aerosol particles, size-resolved
685
hydrometeor particles, radiation, and surface processes, contains arrays with a
minimum of
N
min
= (N
M
+ N
G
+ N
D
N
B
N
V
+ N
R
)N
3D
+ N
S
N
2D
(21.1)
oating-point values, where N
M
is the number of meteorological variables, N
G
is
the number of gases, N
D
is the number of aerosol/hydrometeor size distributions
(assuming aerosol particles can grow into hydrometeor particles), N
B
is the num-
ber of size bins in each size distribution, N
V
is the number of possible chemical
components in particles in each size bin of each distribution, N
R
is the number
of radiative variables, N
3D
is the total number of three-dimensional grid cells in
the model, N
S
is the number of surface variables (e.g., ground temperature, soil
moisture), and N
2D
is the total number of two-dimensional horizontal grid squares
in the model. Whereas, N
min
is the minimum summed array dimension in a model,
additional memory space is needed for work arrays.
Example 21.1
If N
M
= 10. N
G
= 100. N
D
= 5. N
B
= 20. N
V
= 30. N
R
= 2. N
3D
= 50000. N
S
=
6. and N
2D
= 2500. (21.1) implies that N
min
= 156 million array points. Thus,
the model requires over 156 million words of central memory (almost 1.25
gigabytes (GB), since 1 word = 8 bytes) just to store values of variables. If only
meteorology is treated, N
min
= 515 000.
21.1.6 Selecting a computer architecture
The next step in model development is to select a computer architecture for building
the model on. Example architectures are single-processor scalar, single-processor
vector, multiprocessor shared-memory scalar, multiprocessor shared-memory vec-
tor, and multiprocessor distributed-memory scalar architectures.
A scalar processor is a processor that operates on variables in a loop, one at a
time. For example, in the Fortran loop
DO 100 I = 1. 150 (21.2)
DVAR(I) = AVAR(I)
BVAR(I)
100 CONTINUE
150 multiplications are performed in sequence on a scalar processor. A single-
processor scalar machine has one processor performing scalar operations.
Avector processor is a processor that operates on several variables within a loop
at the same time. In loop (21.2), a vector processor may break down variables AVAR
and BVAR each into two registers of 64 and one register of 22 values. Elements
686
fromthe rst register of each AVARand BVARare loaded into a functional unit that
multiplies values of AVAR and BVAR. At least seven values of AVAR and BVAR
can be loaded into the functional unit at the same time. Thus, several operations
are carried out simultaneously on a vector processor. Vector processors operate
only on the inner loop of a nested loop. A single-processor vector machine has
one processor performing vector operations. For a code to run fast on a vector
machine, every inner loop must be optimized for the vector processor. Such an
optimized code is often efcient on a scalar processor as well.
A multiprocessor machine, also called a parallel-processor machine, is one in
which several processors operate on different parts of a program simultaneously.
Separate single-processor machines can be linked together as a network to per-
form the same function as a single multiprocessor machine. On a multiprocessor
machine, each of the 150 multiplications in loop (21.2) can be distributed to a
separate processor. A shared-memory multiprocessor machine is a multiprocessor
machine in which memory is common to all processors. A distributed-memory
multiprocessor machine is a multiprocessor machine in which memory is allocated
to each processor and not shared by all processors.
21.1.7 Coding the model
The next step is to codify the model on the chosen architecture in a manner that
is easy to understand, follow, use, and modify. Although some coding may pre-
exist, new coding is often required. A 3-D gas-phase chemical-transport model
may require an emission algorithm, a chemical solver, a transport module, and a
dry-deposition algorithm. If all but the emission algorithm pre-exist, the modeler
must develop the emission algorithm and link it with the other modules.
Two methods of making the code easy to understand are to include many com-
ments and references. Comments are important, not only for the model developer,
who may need to edit the model after several years, but also for others who may
use and/or modify the model. Comments include denitions, units of variables, and
descriptions of equations. References are important, so that users of the model can
check the origin of equations or numerical algorithms, either to understand them
or to determine whether they should be replaced.
Two methods of making the code easy to followare to use indentation frequently
and to select appropriate variable names. Indenting nested loops (e.g., DO . . .
CONTINUE loops or IF . . . THEN statements in Fortran) makes the code more
readable than left-justifying all coding. Selecting descriptive variable names enables
a user to follow the code easily. The variable names, TEMPK, PRESS, UWIND,
and VWIND are easier to follow than are T, P, U, and V, respectively. Variables
should consist of no fewer than three characters because one- or two-character
names are difcult to search for. A search for all occurrences of T in a subroutine,
for example, may generate a list of thousands of Ts.
A method of making the code easy to use is to control portions of the code by
onoff switches. For example, it may be desirable not to solve gas chemistry for a
sensitivity test. This can be accomplished most readily by putting an IF statement,
687
such as
IF (IFCHEM.EQ.1) CALL SUBROUTA (21.3)
in the model, where IFCHEM= 1 for solving gas chemistry and 0 for ignoring it.
Two methods of making the code easy to modify are to create external datales
for reading variables that are frequently changed and to allow the code to gener-
ate arrays from input data rather than hardwiring the arrays into the code. For
example, if an onoff switch is frequently changed, it is less time consuming to
change the switch within an external data le, which is read by a compiled code
during runtime, than to change the switch within the code, then to recompile and
run the code. In the second example, a chemical solver is easier to modify if its
reactions are read from an external dataset and its rst derivatives and partial
derivatives are calculated automatically in the code rather than if the reactions,
rst derivatives, and partial derivatives are each hardwired in the code. In the for-
mer case, it is not necessary to change anything except the reaction information
when a reaction is changed; in the latter, it is necessary to change up to hundreds of
derivatives and partial derivatives when a reaction is changed, increasing the risk of
error.
21.1.8 Optimizing the model
Another important step in model development is to increase the speed and reduce
the memory requirements of the model. One method of increasing the speed of
individual loops on a given computer is to vectorize the loop (Section 12.8). Another
method is to use sparse-matrix techniques (Section 12.8).
Reducing memory requirements is important when computer memory is lim-
ited. Two methods of reducing memory requirements are to minimize the number
of global arrays and to prevent work arrays from having a dimension equal to
the size of the grid domain. Global arrays are arrays used to store information
continuously over the entire domain. Example global arrays are arrays for temper-
ature, pressure, wind speed, wind direction, relative humidity, gas concentration,
and aerosol component concentration. Work arrays are arrays used temporarily.
Global arrays must have one of their dimensions equal to the size of the grid
domain. Thus, minimizing the number of global arrays is one method of reducing
memory requirements. Although work arrays are often used to store information
from global arrays temporarily, there is no reason that a work array ever has to
have any dimension equal to the size of the grid domain. The reason is that the
model domain can be divided into blocks (groups) of grid cells, where operations
are carried out one block at a time rather than over the whole domain at once
(Section 12.8). In such cases, one dimension of each work array is reduced from
the number of grid cells in the model domain to the number of grid cells in a
block.
A third method of reducing memory requirements is to write global-array infor-
mation to a data le rather than to store the information in a global array. The
688
tradeoff is that additional computer time is required to read and write the data,
and additional disk space is required to store the data.
21.1.9 Time steps and intervals
The next step in model development is to select model time steps and time intervals.
These parameters depend on the desired accuracy, the computer time available, and
stability limitations of the code. For a 5-km 5-km horizontal grid, a typical time
step for hydrostatic dynamical calculations is 5 s. For a 5
global grid, it is
around 300 s. For gas and aqueous chemistry, the time step is variable with some
solution methods and xed with others. The Gear chemical integration method
(Chapter 12) uses a predicted time step that varies from-10
4
to >900 s, depending
on the stiffness of the system of equations. The MIE chemical solver, described in
the same chapter, uses a xed time step of 1, 10, 30, or 60 s. Generally, the longer the
xed time step, the less accurate the solution and the less computer time required.
Cloud, aerosol, radiative, transport, and surface processes use xed or variable time
steps. Some processes, such as coagulation, are slow enough away from sources to
allow time steps of minutes to days, depending on the application.
In Section 6.2, a time interval was dened as the period during which several
time steps of a process are solved without interference by another process. For
time-consuming processes in which the time step is constant (e.g., some aerosol
processes), the time interval often equals the time step. For processes that require
short time steps (e.g., dynamics) or variable time steps (e.g., chemistry), the time
interval often exceeds the time step. In 3-D models, time intervals often range from
1 minute to 1 hour.
21.1.10 Initial conditions
A model calculates output parameters from a set of initial conditions, boundary
conditions, input data, and model equations. This subsection discusses initial con-
ditions. Boundary conditions and input data are discussed in the next two subsec-
tions.
Initial conditions are needed for meteorological, surface, gas, and aerosol vari-
ables in a model. Meteorological variables that are usually initialized in three
dimensions include air temperature, air pressure, specic humidity, wind veloc-
ity, and geopotential height. Surface variables include initial soil moisture and
ground temperature. Gas variables include gas mixing ratio. Aerosol variables
include aerosol number concentration and component mole or volume concen-
tration as a function of particle size.
For short-term simulations of air pollution and weather, accurate initialization
is critical for producing accurate comparisons of model predictions with data. For
long-term simulations of climate, initialization is less important than for short-
term simulations for two reasons. First, the lifetimes of most (but not all) gases and
aerosol particles are much shorter than are simulation times for climate studies,
which are on the order of years to centuries. Second, time-dependent solar and
689
thermal-infrared radiation drive the general circulation of the atmosphere, gradu-
ally diminishing the effect of initial conditions.
Two ways to initialize a model are to interpolate observations and to use out-
put from another model. For simulations of regional air pollution episodes, for
which many pollutant concentrations and meteorological variables are measured
during a eld campaign, initial conditions may be interpolated from measurements
taken at the simulation start time. For global simulations, where data are sparse,
meteorological variables are generally initialized with results froma previous global
simulation. In some cases, the prior simulation may be run for many years to obtain
a long-term-average result and to reduce the effect of initial conditions. In other
cases, the prior simulation may be nudged toward observed values from a sparse
network of soundings and surface stations. This practice is called data assimilation.
A sounding is a vertical prole of variables such as temperature, pressure, specic
humidity, and velocity as a function of height, measured from a balloon released
from the surface. A surface station measures similar variables, but only near the
surface. Data assimilation is performed by relaxing model values toward observed
values in a manner similar to the relaxation technique described for treating bound-
ary conditions in the next subsection.
21.1.11 Boundary conditions
Boundary conditions are needed in all models, including global, hemispheric,
limited-area, and nested models.
In global and hemispheric models, westeast boundary conditions are periodic
(the western edge of the westernmost grid cell is adjacent to the eastern edge of the
easternmost grid cell); thus, complex lateral boundary conditions are not needed.
At the poles in a spherical-coordinate global model, southnorth velocities may be
set to zero so that winds travel westeast around polar singularities.
In limited-area models and at the southern and/or northern boundary of a hemi-
spheric model, lateral horizontal boundary conditions are needed. Some types of
lateral boundary conditions for meteorological variables are described in Chapter 7.
For gases and aerosol particles, many models assume that constant concentrations
are advected into the model domain from outside a lateral boundary. In reality, gas
concentrations outside a boundary vary during the day and night due to chemistry.
One way to address this feature is to dene a virtual row or column outside a
lateral boundary wherein time-dependent chemical equations are solved for gases.
The resulting virtual-boundary concentrations are then advected into the domain.
Virtual boundaries, in which time-dependent physical and chemical processes are
solved, can also be created for aerosol particles.
In a nested model, lateral boundary conditions in the global domain are the
same as those in any other global model. In progeny domains, meteorological
variables are relaxed toward values interpolated from the parent domain (e.g.,
Davies 1976; Kurihara and Bender 1983; Anthes et al. 1989; Giorgi et al. 1993;
Lu and Turco 1994; Marbaix et al. 2003). Relaxing a progeny variable in this
case means allowing it gradually to approach the interpolated parent variable.
690
Relaxing is necessary only if the time interval for passing variables between par-
ent and progeny domains (the nesting time interval, h
nest
) exceeds the time step
of the process that affects the variable in the progeny domain (h). For example,
if variables are passed between the parent and progeny domain every hour, and
the time step for dynamical calculations in the progeny domain is ve seconds,
meteorological variables in the progeny domain may be relaxed toward parent
variables. If the nesting time interval also equals ve seconds, progeny variables
can instead be set directly to interpolated parent variables each time step without
relaxation.
Meteorological variables are generally relaxed in a buffer zone adjacent to each
boundary in the progeny domain. The buffer zone generally consists of the k =
1 . . . Nwesternmost and easternmost columns and southernmost and northernmost
rows of the progeny domain (where N is often set to 5 but may range from 1 to 10,
and k = 1 represents outer-boundary columns and rows). One method of relaxing
is rst to calculate variable values at the end of a nesting time interval h
nest
(s)
in the parent domain. Then, during the mth time step, h (s), of the same nesting
interval, variable values in the buffer zone of the progeny domain are updated
with
prog
t
0
+mh.k
=
par
t
0
+h
nest
.k
+ F
k
prog
t
0
+(m1)h.k
1 + F
k
(21.4)
where t
0
is the time (s) at the beginning of the current nesting interval, t
0
+h
nest
is
the time at the end of the nesting interval, t
0
+(m1)h is the time at the beginning
of a progeny-domain time step during the nesting interval (where m is an integer
between 0 and h
nest
h), t
0
+mh is the time at the end of the progeny-domain time
step,
prog
is a parameter value in the progeny domain,
par
is a value of the same
parameter interpolated from the parent domain to the location of
prog
in the
progeny domain, and
F
k
=
(h
nest
h)
h
[1 e
(1k)M
] (21.5)
is an expression for the relaxation coefcient, where M is a coefcient selected
arbitrarily between 0.5 and 5. Equation (21.5) predicts that, when h
nest
= h, F
k
=
0, and
prog
=
par
for all k = 1 . . . N. When h
nest
> h, F
k
= 0 at the boundary,
forcing
prog
=
par
at the boundary. Because F
k
increases with increasing distance
from the boundary, the relative weight of the parent-domain value decreases with
increasing distance from the boundary in (21.4). The product F
k
h is the e-folding
lifetime for damping at the boundary.
Gas and aerosol concentrations do not need to be relaxed. Instead, values from
the parent domain may be interpolated to a virtual boundary outside the progeny-
domain boundary, and these concentrations may be advected into the progeny
domain.
For nested and nonnested models on all scales, vertical boundary conditions are
needed. At the model top and surface, vertical scalar velocities are usually set to
zero, and variables are not transported through the model top or bottomboundaries
691
by the wind. At the surface, energy and moisture uxes are exchanged between the
atmosphere and soil, vegetation, water, and other surfaces. In addition, gases and
aerosol particles are emitted into the atmosphere. Dry deposition, sedimentation,
and wet deposition uxes remove gases and aerosol particles from the bottom
atmospheric layer. Many global atmospheric models are coupled with a 2-D or
3-D ocean model. In such cases, the bottom boundary of the total model over
the ocean is the bottom of the ocean mixed layer (for a 2-D ocean model) and
the topographical bottom of the ocean (for a 3-D ocean model). Almost all 3-D
atmospheric models that treat meteorology are coupled with a single- or multi-
layer soil model. In such cases, the bottom boundary of the total model over soil
is the bottom soil layer.
21.1.12 Input data
Models include some input data that are xed and other data that vary in space
and time. Topographic data are generally xed in space and time. They are used
to specify geopotential at the surface, which affects the hydrostatic equation in
a sigma-pressure coordinate model (Section 7.6) and the denition of sigma in a
sigma-altitude coordinate model (Section 5.4.1).
Solar radiation drives much of the general circulation of the atmosphere. In
a model, yearly-averaged, top-of-the-atmosphere, spectral solar irradiance data
(Appendix Table B.2) are used as a model input. The yearly-averaged data are
scaled with day of the year (e.g., with the equations in Section 9.8.2), because the
EarthSun distance varies with day of year.
Land-use data (which may vary with month or season) are used to estimate
average grid-cell values of the surface roughness length for momentum, soil specic
heat, soil density, soil porosity, and other parameters. For example, if each grid cell
in a model is subdivided by fractional land-use type, an average surface roughness
length for momentum in the grid cell can be calculated as
z
0.m.c
= f
l.1
z
0.m.1
+ f
l.2
z
0.m.2
+ + f
l.i
z
0.m.i
(21.6)
where z
0.m.i
is the surface roughness length for momentum of land-use type i,
and f
l.i
is the fractional area of the grid cell consisting of land-use type i. Similar
equations can be written for other parameters. The fractions satisfy f
l.1
+ f
l.2
+
+ f
l.i
+ = 1. Surface roughness lengths are used to estimate, among other
parameters, eddy diffusion coefcients and dry-deposition speeds.
Chemical-rate coefcient data are essential for simulating gas, aqueous, or
reversible chemistry. Such data include temperature- and/or pressure-dependent
uni-, bi-, and termolecular rate coefcients for gas- or aqueous-phase chemical
reactions (e.g., Appendix Tables B.4 and B.8) and equilibrium-coefcient data for
reversible reactions (e.g., Appendix Table B.7).
Absorption cross-section data are used to calculate absorption coefcients of
gases and liquids. Absorption coefcients are used for photolysis and optical depth
692
calculations. Some sources of absorption cross-section data are given in Appendix
Table B.4.
Activity-coefcient data are used for simulating reversible chemistry in concen-
trated aerosol particles. Some temperature-dependent activity-coefcient data are
given in Appendix Table B.9.
Emission data are needed to simulate gas and particle pollution buildup. Some
emission rates, such as of soildust particles, sea-spray drops, biogenic gases, and
nitric oxide fromlightning, are often calculated fromphysical principles in a model
(e.g., Chapter 14; Chapter 18), so are not considered input data. Other emission
rates, such as of volcanic, biomass burning, and fossil-fuel combustion sources are
usually read in as a datale. Such datales can be global or regional in space and
may be hourly, daily, monthly, seasonally, or yearly in time.
21.1.13 Ambient data
A models performance is judged by how its predictions compare with data. Data
used for comparing results with are ideally an extension of the data used for ini-
tializing the model. In other words, if time-dependent measurements are available,
data for the time corresponding to the beginning of the model simulation should
be used to initialize the model, and data for all subsequent times should be used to
compare with model predictions.
Model predictions are generally compared with gas, aerosol, cloud, radia-
tive, meteorological, and surface data. Some gas parameters frequently compared
include ozone, carbon monoxide, nitrogen dioxide, water vapor, sulfur dioxide,
total reactive organic gases, and peroxyacetyl nitrate. Aerosol parameters com-
pared include sulfate, nitrate, chloride, ammonium, sodium, black carbon, organic
matter, and soil dust. Cloud parameters compared include cloud optical depth,
cloud fraction, cloud liquid, cloud ice, and precipitation. Radiative parameters
compared include aerosol optical depth, single-scattering albedo, scattering extinc-
tion coefcient, absorption extinction coefcient, surface albedo, solar irradiance,
ultraviolet irradiance, thermal-infrared irradiance, and actinic ux. Meteorolog-
ical parameters compared include air temperature (minimum, maximum, mean),
air velocity, relative humidity, air pressure, and geopotential. Surface parameters
compared include ground temperature, soil moisture, latent heat ux, and sensible
heat ux.
Some models assimilate data (nudge results toward measurements) to improve
their comparisons with data. Nudging is sometimes used when running a model in
the past (hindcasting) in order to create an initial meteorological eld for running
other simulations in the future (Section 21.1.10). Nudging is also used to predict
variables in the future (forecasting) by running a model prognostically for a period,
nudging the prediction toward observations, then running the model prognostically
for another period, and so on. A true forecast model, though, does not use nudging
since, by denition, forecasting is the prediction of future events without knowledge
of the future, and nudging assumes knowledge of the future. Another problemwith
693
O
A
B
C
D
E
r
I O
A
B
C
D
E
(a) (b)
Figure 21.2 (a) Domain of inuence around point O. The letters
A, B, C, D, and E represent locations where data are available
for interpolation to model point O. (b) Division of the domain
of inuence into sectors.
the use of nudging is that it impairs the evaluation of the accuracy of a model. If a
model is nudged, errors are hidden.
21.1.14 Interpolating data and model results
Ambient, input, and output data are often interpolated in a model. Ambient data are
interpolated to model grid cells to initialize the model. Emission data are sometimes
interpolated between coordinate systems. Model predictions are interpolated to
locations where ambient measurements are taken so that model results can be
compared with data. A basic interpolation method is discussed for each of these
cases.
21.1.14.1 Interpolating from scattered points to a xed point
Scattered ambient data are often interpolated horizontally and vertically to each
model grid cell for initialization. Goodin et al. (1979) provide a review of several
interpolation methods. One method for horizontally interpolating scattered data
to a xed point is with inverse-square interpolation, whereby each datum value is
weighed by the inverse square of the distance of the datum location to the point
of interest. Only data within a predened domain of inuence are considered. The
domain of inuence is a circular area of a given radius of inuence. Figure 21.2(a)
shows an example domain of inuence with radius of inuence r
I
, around a point
O. The domain of inuence contains ve locations A, B, C, D, and E where
data values are known. The value at point O is interpolated from the values at the
data locations with
V
O

V
A
d
2
AO
+ V
B
d
2
BO
+ V
C
d
2
CO
+ V
D
d
2
DO
+ V
E
d
2
EO
d
2
AO
+d
2
BO
+d
2
CO
+d
2
DO
+d
2
EO
(21.7)
694
where V
A
, V
B
, . . . , etc. are the known data values at points A, B, . . . , etc., and
d
AO
, d
BO
, . . . , etc. are the distances from points A to O, B to O, . . . , etc.
Advantages of inverse-square interpolation are that it weighs nearby data points
more than distant data points, and it is simple to implement. A disadvantage is that
when two points, such as A and B in Fig. 21.2(a), are close together, they are each
weighed the same as any other point of the same distance from O. A method to
reduce this problem is to multiply each datum value by an angular distance (in
radians) found by drawing a line from point O through the midpoint between each
pair of data locations, as shown in Fig. 21.2(b). The angles (
A
,
B
, . . . , etc.)
are constrained by
A
+
B
+
C
+
D
+
E
= 2. Including the angle-dependence
in (21.7) gives
V
O

A
V
A
d
2
AO
+
B
V
B
d
2
BO
+
C
V
C
d
2
CO
+
D
V
D
d
2
DO
+
E
V
E
d
2
EO
A
d
2
AO
+
B
d
2
BO
+
C
d
2
CO
+
D
d
2
DO
+
E
d
2
EO
(21.8)
With this equation, a datum value at point D now carries approximately the same
weight as the data values at points A and B combined.
21.1.14.2 Interpolating from one coordinate system to another
When data are provided in one coordinate system but a model uses a different
coordinate system, the data must be interpolated between coordinate systems. For
example, emission data for urban modeling are often developed on a Universal
Transverse Mercator (UTM) grid projection, which has rectangular grid cells. A
model may be run on a spherical (geographic) grid projection. When a UTMgrid lies
on top of a geographic grid, UTM cell boundaries cross geographic cell boundaries
at random locations; thus, interpolation is needed between the two grids.
The rst interpolation step is to determine the area of each UTM grid cell that
lies within a geographic grid cell. Since no formulae are available to calculate such
areas, the areas must be obtained by physical integration. Each UTM cell, which
may be 5 km 5 km in area, can be divided into 10 000 or more smaller cells
(subcells), each 50 m 50 m in area. Each subcell is assumed to contain the
same datum value as the larger UTM cell that the subcell lies in. The latitude and
longitude of each subcell corner are found from UTM-to-geographic conversion
formulae (US Department of the Army 1958). Once geographic boundaries of
each subcell are known, the subcells from each large UTM cell U that fall within a
geographic cell Gare counted, and the sumis denoted by N
U,G
. Fromthe sum, two
types of interpolated values can be obtained: cumulative and average. Cumulative
values are those in which all data from the original UTM cells that fall within
the geographic cell of interest are summed. Cumulative values are necessary for
interpolating emission data. Average values are those in which all data from the
original UTM cells that fall within the geographic cell of interest are averaged.
Average values are necessary for interpolating most other data, such as land use
and albedo data.
695
A
D
A
E
A
C
A
B
O
B C
E D
Figure 21.3 Location of point O in a
rectangle with points B, C, D, and E at
the corners.
The cumulative and average interpolated values of a variable in each geograph-
ical cell are determined with
V
G.c
=
M
G
U=1
N
U.G
A
M
V
U
A
U
V
G.a
=
1
A
G
M
G
U=1
(N
U.G
A
M
V
U
) (21.9)
respectively, where M
G
is the number of large UTM cells overlapping part of geo-
graphic cell G, A
M
is the area of a UTM subcell, A
U
is the area of a large UTM cell,
A
G
is the area of a geographic cell, and V
U
is the datum value in the large UTM
cell. This interpolation method can be applied to any two-coordinate systems. The
precision of the method improves with improved resolution of subcells.
21.1.14.3 Interpolating from xed points to a random point
A model generally produces output at regularly spaced locations, such as at the
horizontal center or edge of a grid cell. Observational data, used for comparison
with model results, are located at scattered locations throughout a model grid.
Thus, it is necessary to interpolate model predictions from xed points to loca-
tions where data are measured. If grid cells are roughly rectangular, model values
at data locations can be estimated using bilinear interpolation (e.g., Press et al.
1992).
Figure 21.3 shows the locations of known and unknown variables when bilinear
interpolation is used. In this gure, values at points B, C, D, and E are known and
produced by the model. Model values must be interpolated from these points to
point O, and the resulting value is compared with a datum value at the same
point.
With bilinear interpolation, the value interpolated to point O is
V
O

A
B
V
B
+ A
C
V
C
+ A
D
V
D
+ A
E
V
E
A
B
+ A
C
+ A
D
+ A
E
(21.10)
where V
B
, V
C
, V
D
, and V
E
are model values at points B, C, D, and E, respectively,
and A
B
, A
C
, A
D
, and A
E
are the rectangular areas shown in Fig. 21.3. Equation
(21.10) states that, as location O approaches location B, area A
B
increases in size,
increasing the weight of B on the value at O.
696
21.1.15 Statistics and graphics
Numerical models produce output that is difcult to analyze number by number.
Statistics and graphics are useful for evaluating model performance, nding bugs,
and studying model results.
Common statistical measures are the normalized gross error, normalized bias,
paired peak estimation accuracy, and temporally paired peak estimation accuracy
(e.g., Tesche 1988). The overall normalized gross error (NGE) compares the abso-
lute value of a prediction with an observation, summed and averaged over all
observations and times. Thus,
NGE =
1
N
tim
N
obs
N
tim
j =1
N
obs
i =1
| P
x
i
.t
j
O
x
i
.t
j
|
O
x
i
.t
j
(21.11)
whereN
tim
is the number of times that observations were taken, N
obs
is the number
of observations taken at each time, P is a predicted value, O is an observed value,
x
i
is the location of site i, and t
j
is the jth time of observation. In this equation, all
observations and predictions are paired in space and time; thus, they are compared
at the same time and location. Because low measured values are often uncertain,
(21.11) is applied only when the observation is larger than a threshold, called a
cutoff value. With respect to ozone mixing ratios in urban air, cutoff values are
typically between 0.02 and 0.05 ppmv.
Normalized gross errors can be measured at a single location for all times or at
a single time for all locations. The location-specic NGE is
NGE
x
=
1
N
tim
N
tim
j =1
| P
x.t
j
O
x.t
j
|
O
x.t
j
(21.12)
where x is the location of interest. The time-specic NGE is
NGE
t
=
1
N
obs
N
obs
i =1
| P
x
i
.t
O
x
i
.t
|
O
x
i
.t
(21.13)
where t is the time of interest.
The last two parameters are useful for isolating locations or times of poor
prediction. If the location-specic NGEs at 29 out of 30 observations sites are
25 percent but the NGE is 1000 percent at the 30th site, the overall NGE is
57.5 percent. Thus, the overall NGE does not give complete information about
the model performance. The location-specic NGE, in this case, shows that the
model performance at 29 sites was much better than at the 30th site. With this
information, the modeler can focus efforts on determining why the model perfor-
mance at the 30th site was poor.
Many modeling studies compare monthly, seasonally, or yearly averaged data
at a given location or at many locations with model values averaged in the same
way. The resulting statistic, called an unpaired-in-time, paired-in-space error, has
697
the form
UTPSE =
1
N
obs
N
obs
i =1
N
tim
j =1
P
x
i
.t
j

N
tim
j =1
O
x
i
.t
j
N
tim
j =1
O
x
i
.t
j
(21.14)
Even though an absolute value is used in this equation, the UTPSE is not a true
measure of error because the statistic does not pair prediction with measurements
in both time and space, a necessary condition for the statistic to be a true measure
of error. For example, suppose two values measured at the same location but at
different times are 4 and 2 and the two corresponding model values are 1 and 5.
The UTPSE for this set of statistics is 0, suggesting that the model has no error.
However, the NGE for the same set of statistics is 112.5 percent, giving a true
indication of model error.
Other modeling studies compare spatially and temporally averaged data with
spatially-and-temporally averaged model values. Zonally averaged data for Decem-
berJanuaryFebruary and globally averaged data for the month of June are data
of this type. The statistic that results from this type of comparison is the unpaired-
in-time, unpaired-in-space error, and has the form,
UTUSE =
N
tim
j =1
N
obs
i =1
P
x
i
.t
j

N
tim
j =1
N
obs
i =1
O
x
i
.t
j
N
tim
j =1
N
obs
i =1
O
x
i
.t
j
(21.15)
Like with the UTPSE, the UTUSE is not a real measure of model error.
Another statistical measure is normalized bias (NB),
NB =
1
N
tim
N
obs
N
tim
j =1
N
obs
i =1
P
x
i
.t
j
O
x
i
.t
j
O
x
i
.t
j
(21.16)
The NB indicates whether a parameter is over- or underpredicted, on average, in
comparison with the data. The NB does not indicate whether the model perfor-
mance is accurate.
Ameasure of the magnitude of the spread around the mean value of a distribution
is the variance. The biased variance, or square of the standard deviation of a
distribution, is dened as
2
u
=
1
N
N
i =1
(V
i

V)
2
(21.17)
where N is the total number of data values,

V is the mean of all data values, and
V
i
is the i th data value. The biased variance of the time-specic normalized gross
error is
2
u.NGE
t
=
1
N
obs
N
obs
i =1
| P
x
i
.t
O
x
i
.t
|
O
x
i
.t
NGE
t
2
(21.18)
This value gives the spread of gross errors around the mean gross error taken at a
specic time. The unbiased variance is the same as the biased variance, except that
the summation in the unbiased variance is divided by N
obs
1 instead of by N
obs
.
698
The paired peak accuracy (PPA) identies how well a model predicts a peak
observed parameter value at the time and location of the peak. It is given by
PPA =
P
x.
t
O
x.
t
O
x.
t
(21.19)
where the circumex ( ) indicates that a value is taken at the time and location of
the peak observed value. The temporally paired peak accuracy (TPPA) identies
how well the model predicts the peak observed value at the same time of the peak,
but at any other location. It is
TPPA =
P
x.
t
O
x.
t
O
x.
t
(21.20)
The TPPA is less useful than the PPA in that, even if the TPPA is zero, the location
of the predicted peak may be far from that of the observed peak.
Another method of judging the accuracy of a model is with time-series plots,
which are graphical comparisons of model predictions with data for one parameter
at one location over the period of the simulation. Parameters frequently compared
are gas concentration, particle concentration, temperature, relative humidity, wind
speed, wind direction, and solar radiation, among others. Example time-series plots
are shown in Section 21.2. Like location-specic NGEs, time-series plots are often
better indicators of model performance than are overall NGEs.
A third method of judging the accuracy of a model is with spatial comparisons
of model predictions with data at a given time. Two-dimensional contour plots of
predictions can be laid on top of two-dimensional maps of measurements. Such
comparisons allow a modeler to judge whether predictions at a given time are
similar to or different from measurements. Although the interpretation of these
plots is subjective, they are useful for estimating accuracy and whether serious
aws exist in the model.
The statistical and graphical techniques discussed above are used to judge model
performance when data are available for comparison. Graphical displays are also
useful for discovering bugs. A large and unrealistic perturbation in temperature
in a three-dimensional plot may suggest that a programming bug has infected
the model. Brilliant graphics, though, should not be used to argue the validity or
performance of a model. Without evaluation against data, model results are often
open to criticism.
21.1.16 Simulations
After a model has been developed and input and ambient data have been gathered,
simulations can be run. When a simulation is rst started, it usually does not run
to completion because programming bugs still exist in the program. Debugging
can take hours to weeks, depending on the number and severity of bugs and on
the debugging experience of the programmer. Nevertheless, bugs are usually ironed
out, and a baseline simulation can be performed.
699
Once a program has been debugged, it is ready for a baseline simulation. This
type of simulation includes all model processes and input data, and results from it
are often compared with ambient data. The primary purpose of model development
is to study a scientic or regulatory issue, and the baseline simulation should be
designed with such a study in mind. During a baseline simulation, model predictions
and statistical comparisons with data are often stored and/or printed out.
21.1.17 Sensitivity tests
After the baseline simulation, sensitivity tests are often run to gauge the effect of
different assumptions on model performance. The results of such tests should be
compared with data and results from the baseline simulation.
For regional modeling, common sensitivity tests include testing changes in
boundary conditions, initial conditions, and emissions. One test is to set all inow
gas and aerosol concentrations at horizontal boundaries equal to zero and com-
pare the results with those from the baseline case and with data. Another test is
to set all initial gas and aerosol concentrations to zero. A third test is to adjust
the emission inventory to estimate the effect of possible under- or overprediction
of emission on model results. On a global scale, sensitivity tests for emissions and
initial conditions can also be run.
21.1.18 Improving the model
A modeler may nd that simulation results deteriorate over time because of poor
numerical treatment or physical representations in the model. In such cases, better
numerical algorithms or sets of equations may be needed. Modelers are continually
improving and updating their algorithms.
21.2 EXAMPLE MODEL SIMULATIONS
To demonstrate the steps involved in model design, application, and testing, a set
of urban air pollution simulations is discussed. The purposes of the simulations
were to test the effects of aerosol particles on surface air temperatures and to
test the accuracy of a model against data. The model used included many of the
processes shown in Fig. 1.1. Here, results from an application of the model to an
air pollution episode in the Los Angeles basin for August 2728, 1987 are shown
(Jacobson 1997a,b). In the following subsections, model grids, model variables,
ambient data, emission data, initial conditions, and boundary conditions are briey
discussed before results are analyzed.
21.2.1 Model grid
For the simulations, a nonnested limited-area grid with 55 westeast by 38 south
north grid cells was used. The southwest corner was at 33.06
N latitude and
119.1
W longitude. Grid spacing was 0.05 degrees westeast (about 4.6 km) by
700
21.2 Example model simulations
0.045 degrees southnorth (about 5.0 km). Horizontal spherical coordinates were
used. In the vertical, 20 sigma-pressure coordinate layers were used for meteoro-
logical calculations, and 14 were used for all other calculations. The bottom eight
layers (below 850 hPa about 1.5 km) were the same for all processes. The model
top was set to 250 hPa (about 10.3 km).
21.2.2 Model variables and time steps
Variables solved for in the model included horizontal and vertical scalar velocities,
air pressure, air temperature, relative humidity, 106 gas concentrations, 16 aerosol
size bin number concentrations, and 73 aerosol component volume concentrations
per size bin. The time step for meteorology was 6 s, the time step for transport was
300 s, the time step for chemistry varied from -10
4
to 900 s, the time step for
radiation was 900 s, and the time steps for aerosol processes varied from -10
4
to 900 s. The aerosol species included 18 solids, 24 liquids, and 30 ions, and one
category of residual material. The 16 size bins used ranged from 0.014 to 74 m
in diameter.
21.2.3 Ambient and emissions data
August 2728, 1987 was simulated because the number of available ambient mea-
surements was large and a detailed emission inventory was prepared for this period.
Ambient measurements were available for near-surface mixing ratios of ozone, car-
bon monoxide, nitrogen dioxide, nitric oxide, sulfur dioxide, reactive organic gases,
methane, ammonia, and nitric acid. Surface data were available for temperature,
dew point, relative humidity, sea-level pressure, wind speed, wind direction, and
solar radiation. Aerosol measurements of black carbon, organic carbon, sodium,
chloride, ammonium, nitrate, sulfate, and total aerosol mass in the sub-2.5-mand
sub-10-m size regimes were also available. The aerosol and gas emission inven-
tories extended over a region 325 km eastwest by 180 km northsouth, with a
resolution of 5 km in each direction.
21.2.4 Initial conditions
Initial vertical and horizontal proles of temperature, dewpoint, and pressure were
interpolated from 12 sounding sites in and outside the basin for the early morning
of August 27, 1987. Sea surface temperatures were interpolated each hour from
buoy data. Winds were initialized with zero velocities to ensure mass conservation
and to avoid startup waves near mountain regions. Differential heating and cooling
over spatially varying topography created pressure gradients that forced winds to
generate. The Coriolis force, pressure-gradient force, and turbulent uxes affected
the equations of motion over time. Diabatic heating and heat advection inu-
enced changes in potential temperature, which affected pressure gradients. Initial
gas and aerosol concentration were interpolated from data available at 04:30 on
August 27.
701
Table 21.1 Normalized gross errors (NGEs) and normalized biases (NB) after 44 h for the
baseline simulation
No. of NGE NB No. of NGE NB
Parameter comparisons (%) (%) Parameter comparisons (%) (%)
Mass
10
360 50.1 9.3 Na
10
360 36.0 30.2
Mass
2.5
356 43.9 8.1 Cl
10
190 46.8 16.0
BC
10
356 50.6 16.2 O
3
(g) 571 27.8 6.6
BC
2.5
356 57.5 29.9 SO
2
(g) 339 35.4 24.2
OC
10
352 45.4 0.33 NH
3
(g) 269 69.3 25.6
OC
2.5
352 49.0 44.1 HNO
3
(g) 109 54.6 22.3
NH
4 10
325 45.7 30.2 HCHO(g) 61 45.8 32.9
NH
4 2.5
321 55.2 0.15 Temperature 628 0.63 0.09
SO
4 10
304 26.3 52.3 Rel. hum. 358 21.6 4.2
SO
4 2.5
360 28.4 8.3 Solar rad. 50 7.9 3.0
NO
3 10
360 69.8 3.7
sp
125 43.0 13.8
NO
3 2.5
360 67.8 18.4
ap
255 39.8 16.3
The subscripts 10 and 2.5 indicate the mass of the species that resides in particles -10 m and
-2.5 m in diameter, respectively. Mass is total particle mass, BC is black carbon, OC
is organic carbon, Solar rad. is the surface solar irradiance (W m
2
),
sp
is the extinction
coefcient due to particle scattering, and
ap
is the extinction coefcient due to particle absorption.
Cutoff mixing ratios were 50 ppbv for O
3
(g), 5 ppbv for SO
2
(g), 5 ppbv for HCHO(g), 1 ppbv for
NH
3
(g), and 3 ppbv for HNO
3
(g). Other cutoff levels were 0.5 g m
3
for sub-10-m chloride,
2.0 g m
3
for sub-2.5-m ammonium, 0.02 km
1
for
ap
, 10 W m
2
for solar irradiance, and 0
for all other parameters.
21.2.5 Boundary conditions
Outside the horizontal boundaries, initial gas and aerosol concentrations were
interpolated from data. Photochemical calculations were performed on gas con-
centrations in virtual grid cells outside the boundary to simulate their time variation
during the model run. Aerosol concentrations outside the boundary were xed at
low initial values.
21.2.6 Results from baseline simulation
A baseline simulation was run from 04:30 PST, August 27 to 0:30 PST, August
29, 1987. Table 21.1 shows statistical results from the simulation. The statistics
indicate that the NGEs for sulfate, sodium, light absorption, surface solar radiation,
temperature, relative humidity, sulfur dioxide gas, formaldehyde, and ozone were
the lowest among the parameters compared. The NGEs for nitrate and ammonia
gas were largest.
Figure 21.4 shows time-series comparisons of predictions with data for sev-
eral parameters and locations. In some plots, three curves are shown. The third
curve is a prediction from a sensitivity simulation in which aerosol processes were
turned off in the model. The other two curves are model predictions from the
702
0
50
100
150
200
250
300
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Los Angeles
Total mass < 2.5 m
0
50
100
150
200
250
300
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Claremont
Total mass < 10 m
0
2
4
6
8
10
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Claremont
EC < 2.5 m
0
2
4
6
8
10
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Long Beach City College
EC < 10 m
Black carbon
Organic carbon
Sodium
Chloride
0
5
10
15
20
25
30
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Hawthorne
OC < 2.5 m
0
5
10
15
20
25
30
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Anaheim
OC < 10 m
0
2
4
6
8
10
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Sodium < 10 m
0
2
4
6
8
10
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Riverside
Sodium < 10 m
0
1
2
3
4
5
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Hawthorne
Chloride < 10 m
0
1
2
3
4
5
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Riverside
Chloride < 10 m
Particle mass
Figure 21.4 Time series comparisons of baseline (gas and aerosol processes included) model
predictions (solid lines) with data (short-dashed lines) from 04:30 PST August 27 to 0:30
PST August 29, 1987. In the case of formaldehyde, circles are observed values. Most obser-
vational data were given as an average over a 4-h interval. In such cases, model predictions
were averaged over the same interval. In plots with three curves, the third curves (long-
dashed lines) are the predicted-value curves with gas, but not aerosol processes, turned on.
Ammonia
Ammonium
Sulfate
Light scattering
0
10
20
30
40
50
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Hawthorne
Nitrate < 2.5 m
0
10
20
30
40
50
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Nitrate < 2.5 m
0
0.005
0. 01
0.015
0. 02
0.025
0. 03
0 8 16 24 32 40 48 M
i
x
i
n
g

r
a
t
i
o

(
p
p
m
v
)
Anaheim
NH
3
(g)
0
0.005
0. 01
0.015
0. 02
0.025
0. 03
0 8 16 24 32 40 48 M
i
x
i
n
g

r
a
t
i
o

(
p
p
m
v
)
Claremont
NH
3
(g)
0
5
10
15
20
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Hawthorne
Ammonium < 2.5 m
0
5
10
15
20
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Los Angeles
Ammonium < 10 m
0
5
10
15
20
25
30
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Burbank
Sulfate < 2.5 m
0
5
10
15
20
25
30
0 8 16 24 32 40 48
C
o
n
c
e
n
t
r
a
t
i
o
n

(
g

m
3
)
Los Angeles
Sulfate < 10 m
0
0. 2
0. 4
0. 6
0. 8
1
0 8 16 24 32 40 48
s
p

(
k
m
1
)
Claremont
0
0. 2
0. 4
0. 6
0. 8
1
0 8 16 24 32 40 48
s
p

(
k
m
1
)
Nitrate
Figure 21.4 (cont.)
704
21.2 Example model simulations
Surface solar irradiance
Temperature
Relative humidity
Wind speed
0
0. 05
0. 1
0. 15
0. 2
0 8 16 24 32 40 48
a
p

(
k
m
1
)
Burbank
0
0. 05
0. 1
0. 15
0. 2
0 8 16 24 32 40 48
a
p

(
k
m
1
)
Claremont
0
500
1000
1500
0 8 16 24 32 40 48
S
o
l
a
r

i
r
r
a
d
i
a
n
c
e

(
W

m
2
)
CM44 (Riverside)
0
500
1000
1500
0 8 16 24 32 40 48
S
o
l
a
r

i
r
r
a
d
i
a
n
c
e

(
W
m
2
)
CM62 (Temecula)
0
10
20
30
40
0 8 16 24 32 40 48
T
e
m
p
e
r
a
t
u
r
e

(
C
)
Hawthorne
0
10
20
30
40
0 8 16 24 32 40 48
T
e
m
p
e
r
a
t
u
r
e

(
C
)
Claremont
0
20
40
60
80
100
0 8 16 24 32 40 48
R
e
l
a
t
i
v
e

h
u
m
i
d
t
y

(
%
)
Hawthorne
0
20
40
60
80
100
0 8 16 24 32 40 48
R
e
l
a
t
i
v
e

h
u
m
i
d
t
y

(
%
)
Riverside
0
200
400
600
800
1000
0 8 16 24 32 40 48
W
i
n
d

s
p
e
e
d

(
m

s
1
)
Hawthorne
0
200
400
600
800
1000
0 8 16 24 32 40 48
W
i
n
d

s
p
e
e
d

(
m

s
1
)
Riverside
Light absorption
Figure 21.4 (cont.)
705
Formaldehyde
0
0. 05
0. 1
0. 15
0. 2
0. 25
0. 3
0 8 16 24 32 40 48
M
i
x
i
n
g

r
a
t
i
o

(
p
p
m
v
)
Anaheim
0
0. 05
0. 1
0. 15
0. 2
0. 25
0. 3
0 8 16 24 32 40 48
M
i
x
i
n
g

r
a
t
i
o

(
p
p
m
v
)
Los Angeles
0
0. 01
0. 02
0. 03
0. 04
0. 05
0 8 16 24 32 40 48
M
i
x
i
n
g

r
a
t
i
o

(
p
p
m
v
)
Claremont
0
0. 01
0. 02
0. 03
0. 04
0. 05
0 8 16 24 32 40 48
M
i
x
i
n
g

r
a
t
i
o

(
p
p
m
v
)
Ozone
Figure 21.4 (cont.)
baseline simulation and data, respectively. The ammonia gures, for example, show
that the inclusion of aerosol particles in the model was necessary for properly
predicting ammonia gas mixing ratios.
Other statistics fromthe 44-h simulation indicate that aerosol particles increased
nighttime surface air temperatures by about 0.77 K, decreased daytime tempera-
tures by about 0.08 K, and increased overall temperatures (day plus night) by
0.43 K. Nighttime temperatures increased due to aerosol absorption of infrared
radiation fromthe Earths surface. Daytime temperatures decreased due to reduced
solar radiation from aerosol scattering that exceeded warming due to aerosol
absorption.
21.2.7 Results from sensitivity tests
Two additional sensitivity tests are discussed briey. In the rst, aerosol concen-
trations entering the model domain from all lateral boundaries were set to zero.
In the second, lateral and initial concentrations of both gases and aerosol particles
were set to zero.
The rst sensitivity test was run to test the reliability of the boundary condi-
tions. The test demonstrated that setting particle concentrations outside the lateral
boundaries to zero did not change results signicantly for most species, in part
because most comparisons with data occurred far from the boundaries. The aver-
age change in model error was less than 1 percent.
706
The second sensitivity test was run to test the effect of initial conditions on
results. When initial concentrations of gases and aerosol particles were set to zero,
model accuracy degraded for most pollutants. Ozone gross errors increased by over
50 percent, ne sulfate prediction errors doubled, and errors for several other par-
ticle components increased. The removal of initial values turned overpredictions
into underpredictions for some species, but degraded prediction results for most.
Thus, initializing concentrations with realistic values was important for maintain-
ing accuracy during the two-day simulation.
21.3 SUMMARY
In this chapter, model design, application, and testing were discussed. Important
steps in developing and applying a model are to dene the purpose of the model,
select appropriate algorithms for simulation, obtain sufcient input data, and com-
pare model results with data. Other steps include determining an appropriate com-
puter architecture, optimizing the computer code, setting initial and boundary
conditions, running sensitivity tests, improving numerics, and improving physi-
cal parameterizations. Some of the most useful methods of comparing results with
data include calculating normalized gross errors and developing time-series plot
comparisons of predictions with observations. Simulations of a pollution episode
indicated that a model can reasonably simulate atmospheric processes so long as
data, such as initial and emission data, are available.
21.4 PROBLEMS
21.1 Interpolate data values 3, 7, 11, 4, and 6, located at geographic points
(119.2
W, 32.5
N), (119.4
W, 32.6
N), (118.9
W, 32.3
N), (119.0
W,
32.4
N), and (119.1
W, 32.3
N), respectively, to the location (119.3
W,
32.4
N) with the inverse-square method. What are the main advantage and
the main disadvantage of this method?
21.2 Interpolate data values 4, 8, 12, and 3, located at geographic points
(117.3
W, 34.5
N), (117.3
W, 34.6
N), (116.9
W, 34.6
N), and
(116.9
W, 34.5
N), respectively, to the location (117.2
W, 34.53
N) with
bilinear interpolation. For purposes of calculating areas, assume rectangular
shapes. Interpolate the data with the inverse-square method. Discuss differ-
ences in results, if any.
21.3 Write a computer script to calculate the time-specic normalized gross error,
time-specic normalized bias, variance, standard deviation, peak prediction
accuracy, and temporally paired peak prediction accuracy of the data in
Table 21.2.
707
Table 21.2 Predicted and observed O
3
mixing ratios at 17 stations at 10:30 a.m.
Station A B C D E F G H I J K L M N O P Q
Prediction 7.2 8.4 9.6 8.2 6.5 4.3 3.2 3.8 4.8 5.9 6.1 7.0 7.7 8.2 9.2 8.7 7.3
Observation 6.1 7.2 8.3 8.5 7.4 5.8 4.9 4.8 3.5 5.2 7.2 8.1 5.5 5.3 6.3 7.1 5.1
Mixing ratios are in units of parts per hundred million by volume.
708
A P P E N D I X A
Conversions and constants
Syst` eme Internationale (SI) and centimeter-gram-second (CGS) units are used.
Table A.1 lists SI units and their conversion to CGS units. The conversions after
the table include conversions to English units as well.
Table A.1 Syst` eme Internationale (SI) and
Centimeter-Gram-Second (CGS) Units
Quantity SI base or derived unit CGS unit equivalent
Length meter (m) 10
2
cm (centimeters)
Mass kilogram (kg) 10
3
g (grams)
Temperature kelvin (K) K
Time second (s) s
Force newton (N) 10
5
dyn (dynes)
Pressure pascal (Pa) 10 dyn cm
2
Energy joule (J) 10
7
erg (erg)
Power watt (W) 10
7
erg s
1
A.1 DISTANCE CONVERSIONS
1 m = 100 cm = 1000 mm = 10
6
m
= 10
9
nm = 10
10
= 0.001 km
= 39.370 in = 3.2808 ft = 1.0936 yd = 6.2138 10
4
mile
A.2 VOLUME CONVERSIONS
1 m
3
= 1000 L = 10
6
cm
3
= 10
18
m
3
= 264.172 US gallon = 35.313 ft
3
A.3 MASS CONVERSIONS
1 kg = 1000 g = 10
6
mg = 10
9
g
= 10
12
ng = 0.001 tonne (metric) = 0.001 102 3 short ton
= 2.204 62 lb = 35.2739 oz = 6.022 10
26
amu
709
Appendix A Conversions and constants
1 tonne = 1000 kg = 10
6
g = 1.1023 short ton
= 2204.623 lb
1 Tg = 10
12
g = 10
9
kg = 10
6
Mg (megagram)
= 1000 Gg (gigagram) = 0.001 Pg (petagram) = 10
6
tonne (metric)
= 1 Mt (megatonne) = 0.001 Gt (gigaton)
A.4 TEMPERATURE CONVERSIONS
C = K 273.15 = (
F 32)1.8
A.5 FORCE CONVERSIONS
1 N = 1 kg m s
2
= 10
5
g cm s
2
= 10
5
dyn
= 0.2248 lbf (pound-force)
A.6 PRESSURE CONVERSIONS
1 bar = 10
3
mb = 0.986 923 atm = 10
5
N m
2
= 10
5
J m
3
= 10
5
Pa = 10
3
hPa
= 10
5
kg m
1
s
2
= 10
6
dyn cm
2
= 10
6
g cm
1
s
2
= 750.06 torr = 750.06 mm Hg
1 atm = 1.01325 bar = 760 torr = 760 mm Hg
= 29.92 in Hg = 1013.25 hPa = 14.696 psi (lbf in
2
)
A.7 ENERGY CONVERSIONS
1 J = 1 N m = 10
7
erg = 1 W s
= 10
4
mb cm
3
= 10
7
dyn cm = 0.238 902 cal
= 1 kg m
2
s
2
= 10
7
g cm
2
s
2
= 10
5
bar m
3
= 10
4
hPa cm
3
= 6.2415 10
18
eV = 1 C V
= 2.77778 10
7
kWh = 0.009869 L atm = 0.7373 lbf ft
= 9.4782 10
4
Btu
A.8 POWER CONVERSIONS
1 W = 1 J s
1
= 3.41252 Btu hr
1
= 0.013407 horsepower
A.9 SPEED CONVERSIONS
1 m s
1
= 100 cm s
1
= 3.6 km h
1
= 1.94384 knots
= 2.23694 mi hr
1
710
A.10 CONSTANTS
Symbol Quantity Value
A Avogadros number 6.0221367 10
23
molec. mol
1
c speed of light 2.99792 10
8
m s
1
c
p,d
specic heat of dry air at constant
pressure
1004.67 J kg
1
K
1
1.00467 J g
1
K
1
1004.67 m
2
s
2
K
1
0.240 cal g
1
K
1
c
v,d
specic heat of dry air at constant
volume
717.63 J kg
1
K
1
0.718 J g
1
K
1
717.63 m
2
s
2
K
1
0.171 cal g
1
K
1
c
p,V
specic heat of water vapor
at constant pressure (298.15 K) 1865.1 J kg
1
K
1
c
v,V
specic heat of water vapor
at constant volume (298.15 K) 1403.2 J kg
1
K
1
c
W
specic heat of liquid water 4185.5 J kg
1
K
1
4.1855 J g
1
K
1
4185.5 m
2
s
2
K
1
1.00 cal g
1
K
1
F
s
mean solar constant 1365 W m
2
g effective gravity at surface of Earth 9.80665 m s
2
980.665 cm s
2
32.17403 ft s
2
G universal gravitational constant 6.6720 10
11
m
3
kg
1
s
2
6.6720 10
8
cm
3
g
1
s
2
6.6720 10
11
N m
2
kg
2
h Plancks constant 6.6260755 10
34
J s
k von K arm an constant 0.40 (0.350.41)
k
B
Boltzmanns constant (R
A) 1.380658 10
23
J K
1
1.380658 10
23
kg m
2
s
2
K
1
molec.
1
1.380658 10
16
g cm
2
s
2
K
1
molec.
1
1.380658 10
19
cm
3
hPa K
1
molec.
1
1.362603 10
22
cm
3
atm K
1
molec.
1
1.362603 10
22
cm
3
atm K
1
3.298419 10
24
cal K
1
(cont.)
711
1.362603 10
25
L atm K
1
molec.
1
1.380658 10
25
m
3
hPa K
1
molec.
1
L
p
luminosity of the Suns
photosphere
3.9 10
26
W
m
d
molecular weight of dry air 28.966 g mol
1
m
v
molecular weight of water 18.02 g mol
1
M mass of an air molecule (m

d
A) 4.8096 10
26
kg
4.8096 10
23
g
M
e
mass of the Earth 5.98 10
24
kg
5.98 10
27
g
R
universal gas constant 8.31451 J mol

1
K
1
8.31451 kg m
2
s
2
mol
1
K
1
8.31451 10
7
g cm
2
s
2
mol
1
K
1
8.31451 10
4
cm
3
hPa mol
1
K
1
82.06 cm
3
atm mol
1
K
1
0.0831451 m
3
hPa mol
1
K
1
0.08206 L atm mol
1
K
1
8.31451 10
7
erg mol
1
K
1
1.98635 cal mol
1
K
1
R
gas constant for dry air (R
m
d
) 287.04 J kg
1
K
1
0.28704 J g
1
K
1
2.8704 m
3
hPa kg
1
K
1
2870.4 cm
3
hPa g
1
K
1
287.04 m
2
s
2
K
1
2.8704 10
6
cm
2
s
2
K
1
R
e
radius of the Earth 6.371 10
6
m
R
p
radius of the Sun 6.96 10
8
m
R
es
mean EarthSun distance 1.5 10
11
m
R
v
gas constant for water vapor
(R
m
v
)
461.40 J kg
1
K
1
0.46140 J g
1
K
1
4.6140 m
3
hPa kg
1
K
1
4614.0 cm
3
hPa g
1
K
1
461.40 m
2
s
2
K
1
4.6140 10
6
cm
2
s
2
K
1
T
i,f
freezing point of seawater at
35 ppth salinity
271.23 K
T
i,m
melting point of ice 273.05 K
(cont.)
712
T
p
temperature of the Suns photosphere 5796 K
T
s,m
melting point of snow 273.15 K
c
p,d
c
v,d
1.4
0
permittivity of free space 8.85419 10
12
C
2
N
1
m
2
R
c
p,d
0.286
i
thermal conductivity of ice 2.20 W m
1
K
1
sn
thermal conductivity of snow 0.08 W m
1
K
1
B
StefanBoltzmann constant 5.67051 10
8
W m
2
K
4
O angular speed of the Earth 7.292 10
5
s
1
713
A P P E N D I X B
Tables
B.1 STANDARD ATMOSPHERIC VARIABLES
VERSUS ALTITUDE
Table B.1 Variation of gravitational acceleration, air pressure, air temperature, and air
density with altitude in a standard atmosphere
a
Alt.
(km)
Gravity
(ms
2
)
Press.
(hPa)
Temp.
(K)
Density
(kg m
3
)
Alt.
(km)
Gravity
(m s
2
)
Press.
(hPa)
Temp.
(K)
Density
(kg m
3
)
0 9.8072 1013.25 288.15 1.225 22 9.7396 40.5 218.57 0.0645
0.1 9.8069 1001.20 287.50 1.213 23 9.7365 34.7 219.57 0.0550
0.2 9.8066 989.45 286.85 1.202 24 9.7334 29.7 220.56 0.0469
0.3 9.8062 977.72 286.20 1.190 25 9.7304 25.5 221.55 0.0401
0.4 9.8059 966.11 285.55 1.179 26 9.7273 21.9 222.54 0.0343
0.5 9.8056 954.61 284.90 1.167 27 9.7243 18.8 223.54 0.0293
0.6 9.8053 943.22 284.25 1.156 28 9.7212 16.2 224.53 0.0251
0.7 9.8050 931.94 283.60 1.145 29 9.7182 13.9 225.52 0.0215
0.8 9.8047 920.77 282.95 1.134 30 9.7151 12.0 226.51 0.0184
0.9 9.8044 909.71 282.30 1.123 31 9.7121 10.3 227.50 0.0158
1 9.8041 898.80 281.65 1.112 32 9.7091 8.89 228.49 0.0136
1.5 9.8025 845.59 278.40 1.058 33 9.7060 7.67 230.97 0.0116
2 9.8010 795.0 275.15 1.007 34 9.7030 6.63 233.74 0.00989
2.5 9.7995 746.9 271.91 0.957 35 9.6999 5.75 236.51 0.00846
3 9.7979 701.2 268.66 0.909 36 9.6969 4.99 239.28 0.00726
3.5 9.7964 657.8 265.41 0.863 37 9.6939 4.33 242.05 0.00624
4 9.7948 616.6 262.17 0.819 38 9.6908 3.77 244.82 0.00537
4.5 9.7933 577.5 258.92 0.777 39 9.6878 3.29 247.58 0.00463
5 9.7917 540.5 255.68 0.736 40 9.6848 2.87 250.35 0.00400
5.5 9.7902 505.4 252.43 0.697 41 9.6817 2.51 253.11 0.00346
6 9.7887 472.2 249.19 0.660 42 9.6787 2.20 255.88 0.00299
6.5 9.7871 440.7 245.94 0.624 43 9.6757 1.93 258.64 0.00260
7 9.7856 411.1 242.70 0.590 44 9.6726 1.69 261.40 0.00226
7.5 9.7840 383.0 239.46 0.557 45 9.6696 1.49 264.16 0.00197
8 9.7825 356.5 236.22 0.526 46 9.6666 1.31 266.93 0.00171
8.5 9.7810 331.5 232.97 0.496 47 9.6636 1.16 269.68 0.0015
9 9.7794 308.0 229.73 0.467 48 9.6605 1.02 270.65 0.00132
9.5 9.7779 285.8 226.49 0.440 49 9.6575 0.903 270.65 0.00116
10 9.7764 265.0 223.25 0.414 50 9.6545 0.798 270.65 0.00103
11 9.7733 227.0 216.78 0.365 55 9.6394 0.425 260.77 5.7 10
4
12 9.7702 194.0 216.65 0.312 60 9.6244 0.220 247.02 3.1 10
4
13 9.7671 165.8 216.65 0.267 65 9.6094 0.109 233.29 1.6 10
4
14 9.7641 141.7 216.65 0.228 70 9.5944 0.0522 219.59 8.3 10
5
714
B.2 Solar irradiance at the top of the atmosphere
Table B.1 (cont.)
Alt.
(km)
Gravity
(ms
2
)
Press.
(hPa)
Temp.
(K)
Density
(kg m
3
)
Alt.
(km)
Gravity
(m s
2
)
Press.
(hPa)
Temp.
(K)
Density
(kg m
3
)
15 9.7610 121.1 216.65 0.195 75 9.5795 0.0239 208.40 4.0 10
5
16 9.7579 103.5 216.65 0.166 80 9.5646 0.0105 198.64 1.8 10
5
17 9.7549 88.5 216.65 0.142 85 9.5497 0.0045 188.89 8.2 10
6
18 9.7518 75.7 216.65 0.122 90 9.5349 0.0018 186.87 3.4 10
6
19 9.7487 64.7 216.65 0.104 95 9.5201 0.00076 188.42 7.5 10
7
20 9.7457 55.3 216.65 0.0889 100 9.5054 0.00032 195.08 5.6 10
7
21 9.7426 47.3 217.58 0.0757
a
Source: NOAA (1976), except gravity was calculated from (4.40) and (4.45), integrated globally.
The globally averaged effective gravity at the surface of the Earth is 9.8060 ms
2
and occurs at
0.231 km, the globally averaged topographical altitude above sea level.
B.2 SOLAR IRRADIANCE AT THE TOP OF THE ATMOSPHERE
Table B.2 Extraterrestrial solar irradiance, F
s.
(Wm
2
m
1
) at the top of the Earths
atmosphere versus wavelength, (m)
F
s.
F
s.
F
s.
F
s.
F
s.
0.105 0.055 0.355 1125 0.605 1773 0.855 909 3.1 26
0.110 0.050 0.360 1077 0.610 1722 0.860 953 3.2 22.6
0.115 0.039 0.365 1274 0.615 1671 0.865 896 3.3 19.2
0.120 1.168 0.370 1359 0.620 1721 0.870 933 3.4 16.6
0.125 0.371 0.375 1219 0.625 1665 0.875 928 3.5 14.6
0.130 0.060 0.380 1340 0.630 1658 0.880 907 3.6 13.5
0.135 0.080 0.385 1113 0.635 1639 0.885 904 3.7 12.3
0.140 0.061 0.390 1345 0.640 1632 0.890 894 3.8 11.1
0.145 0.063 0.395 1096 0.645 1601 0.895 892 3.9 10.3
0.150 0.096 0.400 1796 0.650 1557 0.9 891 4 9.5
0.155 0.194 0.405 1643 0.655 1502 0.91 880 4.1 8.7
0.160 0.206 0.410 1768 0.660 1562 0.92 869 4.2 7.8
0.165 0.372 0.415 1810 0.665 1570 0.93 858 4.3 7.1
0.170 0.607 0.420 1760 0.670 1539 0.94 847 4.4 6.5
0.175 0.885 0.425 1719 0.675 1556 0.95 837 4.5 5.92
0.180 1.90 0.430 1615 0.680 1526 0.96 820 4.6 5.35
0.185 2.53 0.435 1798 0.685 1481 0.97 803 4.7 4.86
0.190 3.88 0.440 1829 0.690 1460 0.98 785 4.8 4.47
0.195 5.35 0.445 1951 0.695 1491 0.99 767 4.9 4.11
0.200 7.45 0.450 2048 0.700 1453 1.0 748 5 3.79
0.205 10.7 0.455 2043 0.705 1420 1.05 668 6 1.82
0.210 23.4 0.460 2054 0.710 1407 1.1 593 7 0.99
0.215 36.3 0.465 2012 0.715 1376 1.15 535 8 0.585
0.220 44.7 0.470 2007 0.720 1351 1.2 485 9 0.367
0.225 55.0 0.475 2042 0.725 1358 1.25 438 10 0.241
0.230 50.5 0.480 2061 0.730 1331 1.3 397 11 0.165
0.235 49.5 0.485 1867 0.735 1322 1.35 358 12 0.117
0.240 47.0 0.490 1943 0.740 1282 1.4 337 13 0.0851
(cont.)
715
Appendix B Tables
Table B.2 (cont.)
F
s.
F
s.
F
s.
F
s.
F
s.
0.245 62.2 0.495 1993 0.745 1276 1.45 312 14 0.0634
0.250 55.2 0.500 1892 0.750 1272 1.5 288 15 0.0481
0.255 69.5 0.505 1941 0.755 1262 1.55 267 16 0.0371
0.260 111 0.510 1937 0.760 1241 1.6 245 17 0.0291
0.265 212 0.515 1805 0.765 1220 1.65 223 18 0.0231
0.270 255 0.520 1811 0.770 1195 1.7 202 19 0.0186
0.275 197 0.525 1850 0.775 1179 1.75 180 20 0.0152
0.280 186 0.530 1907 0.780 1189 1.8 159 25 0.00617
0.285 317 0.535 1894 0.785 1183 1.85 142 30 0.00297
0.290 546 0.540 1840 0.790 1151 1.9 126 35 0.0016
0.295 573 0.545 1866 0.795 1142 1.95 114 40 0.000942
0.300 493 0.550 1845 0.800 1126 2 103 50 0.000391
0.305 669 0.555 1854 0.805 1112 2.1 90 60 0.00019
0.310 711 0.560 1801 0.810 1080 2.2 79 80 0.0000416
0.315 765 0.565 1828 0.815 1073 2.3 69 100 0.0000257
0.320 777 0.570 1824 0.820 1049 2.4 62 120 0.0000126
0.325 935 0.575 1851 0.825 1050 2.5 55 150 0.00000523
0.330 1041 0.580 1833 0.830 1027 2.6 48 200 0.00000169
0.335 950 0.585 1838 0.835 1012 2.7 43 250 0.0000007
0.340 1035 0.590 1760 0.840 1006 2.8 39 300 0.00000023
0.345 980 0.595 1791 0.845 983 2.9 35 400 0.00000011
0.350 1019 0.600 1752 0.850 951 3 31 1000 0
is the midpoint of a wavelength interval, and the irradiance at each is integrated from the lower
to the upper edge of the interval. For example, the irradiance in the interval from = 0.1475 to
0.1525 m, centered at = 0.15 m is

F
s.
L = 0.096 W m
2
m
1
0.005 m = 0.000 48 W
m
2
. The sum of irradiance over all bins is the solar constant.
The data were derived from Woods et al. (1996) for - 0.275 m, Nicolet (1989) for 0.275 -
- 0.9 m, and Thekaekara (1974) for > 0.9 m.
B.3 CHEMICAL SYMBOLS AND STRUCTURES OF GASES
Table B.3 Inorganic and organic gases and their possible chemical structures
Chemical name
Molecular formula
Chemical structure
Inorganic
Hydrogen and oxygen species
Atomic hydrogen Atomic oxygen (triplet) Molecular oxygen Ozone
H O O
2
O
3
H
O
O O
|
O
O
O
716
B.3 Chemical symbols and structures of gases
Table B.3 (cont.)
Chemical name
Molecular formula
Chemical structure
Inorganic
Hydroxyl radical Hydroperoxy radical Hydrogen peroxide Water vapor
OH HO
2
H
2
O
2
H
2
O
OH
O
H
|
|
O
O
H
|
|
O
|
H
H
|
O
H
|
Nitrogen species
Molecular nitrogen Nitric oxide Nitrogen dioxide Nitrate radical
N
2
NO NO
2
NO
3
N N
N=O
O O
N |
+
|
|
O
O
O
N
+
Nitrous acid Nitric acid Peroxynitric acid Nitrous oxide
HONO HNO
3
HO
2
NO
2
N
2
O
|
O
H
N
O
|
|
+
|
O
H
N
O
|
O
O
O
+
|
|
O H
N O
|
|
O
+
N N
Dinitrogen pentoxide
N
2
O
5
N
O
+
+
| |
|
|
O
O
O O
N
Sulfur species
Atomic sulfur Sulfur monoxide Sulfur dioxide Sulfur trioxide
S (sulfonyl or thionyl radical) SO
2
SO
3
SO
S
S
O O
S
O O
O
S O
Bisulte Sulfuric acid Carbonyl sulde Carbon monosulde
HSO
3
H
2
SO
4
OCS CS
S
O
|
|
HO
O
S
O
| | | | | |
|
OH
O OH C O S
S C
+
(cont.)
717
Appendix B Tables
Table B.3 (cont.)
Chemical name
Molecular formula
Chemical structure
Inorganic
Sulfur species
Carbon disulde Hydrogen sulde radical Hydrogen sulde Methanethiol
CS
2
HS H
2
S (Methyl sulde)
MeSH
C S S
|S
H
|
|
S
H H
CH
3
SH
H
|
|
| |
C
H
S
H
H
|
Methanethiolate radical Methanethiolate oxy Methanethiolate peroxy Methanesulfenic acid
CH
3
S radical radical CH
3
SO
2
CH
3
SOH
CH
3
SO
H
|
| |
C S
H
H
|
H O
|
| | |
C S
H
H
|
H O
|
|
| | |
C
O
S
H
H
|
H O
|
|
| |
C
H
S
H
H
|
Dimethyl sulde Dimethyl sulde radical Dimethyl suldeOH Dimethyl sulfone
DMS CH
3
SCH
2
adduct DMSO
2
CH
3
SCH
3
CH
3
S(OH)CH
3
CH
3
S(O)
2
CH
3
H C
| |
|
| |
C
H
H
S
H
H
|
H C
| |
|
| | | |
C
H
H
S
H
H
|
H H C
| |
|
| |
C
H O
H
S

H
H
| |
|
H
H
H C
| |
|
| |
C
H
O
H
S

H
H
|
H
O
Dimethyl disulde Methane sulfonic acid Hydroxymethane
DMDS MSA sulfonic acid
CH
3
SSCH
3
CH
3
S(O)
2
OH HMSA
HOCH
2
S(O)
2
OH
H
|
|
|
C S
H
H
|
H
|
|
|
C S
H
H
|
H
|
| |
C
O
S
H
H
|
O
OH
HO
|
| |
C
O
S
H
H
|
O
OH
Chlorine species
Atomic chlorine Molecular chlorine Hydrochloric acid Chlorine monoxide
Cl Cl
2
HCl ClO
Cl
ClCl HCl Cl
O
Hypochlorous acid Chlorine peroxy radical Chlorine peroxy radical Dichlorine dioxide
HOCl OClO ClOO Cl
2
O
2
|
O
H
Cl
|
|
|
O O
Cl |
O
O
Cl
|
|
|
O
O
Cl
Cl
718
Table B.3 (cont.)
Chemical name
Molecular formula
Chemical structure
Inorganic
Chlorine species
Chlorine nitrate Chlorine nitrite Methyl chloride Methyl chloroform
ClONO
2
ClNO
2
CH
3
Cl CH
3
CCl
3
N
|
+
|
|
O

O
O
Cl
N
|
+
|
O
O
Cl

H
|
| |
C
H
H
|
Cl H
|
| | |
C
H
H
|
Cl
|
C
Cl
Cl
|
Trichlorouoromethane Dichlorodiuoromethane 1-Fluorodichloro, Chlorodiuoromethane
(CFC-11) (CFC-12) 2-diuorochloroethane (HCFC-22)
CFCl
3
CF
2
Cl
2
(CFC-113) CF
2
ClH
CFCl
2
CF
2
Cl
F
|
| |
C Cl
Cl
Cl
|
F
|
|
|
C Cl
F
Cl
|
F
|
| |
C H
F
Cl
|
Cl
|
| |
C
F
F
|
F
|
|
C
Cl
Cl
|
Carbon tetrachloride
CCl
4
Cl
|
| |
C Cl
Cl
Cl
|
Bromine species
Atomic bromine Molecular bromine Hydrobromic acid Bromine chloride
Br Br
2
HBr BrCl
Br
Br-Br
|
Br H
|
Cl Br
Hypobromous acid Bromine monoxide Methyl bromide Bromine nitrate
HOBr BrO CH
3
Br BrONO
2
|
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | |
Br
H
O
Br-

O H
|
| |
C
H
H
|
Br
N
|
+
|
|
O

O
O
Br
Inorganic carbon species
Carbon monoxide Carbon dioxide
CO CO
2
+
C O
C O
O
(cont.)
719
Appendix B Tables
Table B.3 (cont.)
Chemical name
Molecular formula
Chemical structure
Organic
Alkanes
Methane Ethane Propane Butane
CH
4
C
2
H
6
C
3
H
8
C
4
H
10
C
|
|
H
H
H
H
|
H
|
| |
C
H
H
|
|
C
H
H
|

H
|
H
|
| |
C
H
H
|
|
C
H
H
|

H
|
C
H
H
|
|
H
|
| |
C
H
H
|
|
C
H
H
|

H
|
C
H
H
|
|
|
C
H
H
|
2,2-Dimethylpropane 2,2,4-Trimethylpentane
(neopentane) (isooctane)
C(CH
3
)
4
(CH
3
)
2
CHCH
2
C(CH
3
)
3
H
|
| |
C
CH
CH
|
CH
3
3
3
3
C H
|
| |
C
H
CH
C
2
3
3
C
|
| |
C
CH
CH
|
CH
3
3
3
H
Cycloalkanes
Cyclopropane Cyclobutane Cyclopentane
(CH
2
)
3
(CH
2
)
4
(CH
2
)
5
C
|
H
C
|
C H H
2
2 2
|
C
| |
C
|
H H
|
2 2
C H
2
CH
2
C
|
|
H
C
|
|
C
H
H
2
2
2
C H
2
|
H
C
2
Alkenes
Ethene Propene cis-2-Butene trans-2-Butene
(ethylene) (propylene) C
4
H
8
C
4
H
8
C
2
H
4
C
3
H
6
C
C
|
|
|
H
C
C
H H
3 3
H
|
C
C
|
|
|
H C
C
H
H
3
3
H
|
C
|
H
|
|
H H
C
H
|
C
|
C
C H
H
2
3
H
|
Cycloalkenes
Cyclopentene Cyclohexene 1-Methylcyclohexene
C
5
H
8
C
6
H
10
C
7
H
12
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
CH
3
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
720
Table B.3 (cont.)
Chemical name
Molecular formula
Chemical structure
Organic
Alkynes
Ethyne (acetylene) Propyne
C
2
H
2
C
3
H
4
C H
H C
|
H C
H
|
| |
C
H
H
|
C
Aromatics
Benzene Toluene o-Xylene m-Xylene
C
6
H
6
(methylbenzene) (1,2-dimethylbenzene) (1,3-dimethylbenzene)
TOL XYL XYL
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
C
6
H
5
CH
3
1,2-(CH
3
)
2
C
6
H
4
1,3-(CH
3
)
2
C
6
H
4
CH
3
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
|
|
CH
3
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
CH
3
|
|
|
|
CH
3
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
CH
3
|
|
|
|
Ethylbenzene 1,2,3-Trimethylbenzene
C
6
H
5
C
2
H
5
1,2,3-(CH
3
)
3
C
6
H
3
C
C
H
H
3
3
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
|
|
CH
3
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
CH
3
CH
3
|
|
|
|
|
|
Terpenes
Isoprene d-Limonene -Pinene
(2-methyl-1,3-butadiene) C
10
H
16
C
10
H
16
ISOP
C
5
H
8
|
|
|
C C
|
C
H
3
C
2
|
H
|
|
|
CH
3
C H
2
CH
2
CH
CH
C H
2
|
|
C
C
H
C
|
C
H
3
C
3
|
|
|
H
|
|
|
H
2
C CH
2
CH
C H
C H
3
|
|
C
H
C
| CH
3
C H
2
CH
2
(cont.)
721
Appendix B Tables
Table B.3 (cont.)
Chemical name
Molecular formula
Chemical structure
Organic
Alcohols
Methanol Ethanol o-Cresol Phenol
(methyl alcohol) (ethyl alcohol) CRES (hydroxybenzene)
CH
3
OH C
2
H
5
OH 2-CH
3
C
6
H
4
OH C
6
H
5
OH
H
H
O H
|
| |
C
H
|
|
|
H
H
O
H
|
|
|
C
H
|
|
H
|
C
H
|
C
OH
H
3
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
|
|
|
|
OH
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
|
|
Aldehydes
Formaldehyde Acetaldehyde Propionaldehyde Benzaldehyde
(methanal) (ethanal) (propanal) BZA
HCHO CH
3
CHO CH
3
CH
2
CHO C
6
H
5
CHO
O
H
|
C
H
|
O
H
H
|
|
C
H
|
|
C
H
|
C
O
H
H
3
C CH
2
|
|
O
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
CH
|
|
Glycol aldehyde Acrolein Methacrolein
HOCH
2
CHO (propenal) (2-methyl-2propenal)
CH
2
CHCHO MACR
OH
O
|
|
C |
H
C
H
|
H
|
CH
2
=C(CH
3
)CHO
C
O
H
H
2
C CH
|
C
O
H
H
2
C
|
CH
3
|
C
Ketones
Acetone Methylethylketone Methylvinylketone
(2-propanone) (2-butanone) (3-buten-2-one)
CH
3
COCH
3
CH
3
CH
2
COCH
3
MVK
CH
2
=CHCOCH
3
O
C H
|
|
|
C
H
H
|
H
|
|
|
C
H
H
|
O
H
C
|
|
CH
3
2
C H
3
|
C
O
H
C
|
CH
3
C H
2
|
C
Carboxylic acids
Formic acid Acetic acid Propionic acid
(methanoic acid) (ethanoic acid) (propanoic acid)
HCOOH CH
3
COOH CH
3
CH
2
COOH
C O
HO
H
|
| C O
HO
C H
3
|
|
|
C O
C C H
3 2
HO
H
|
|
722
Table B.3 (cont.)
Chemical name
Molecular formula
Chemical structure
Organic
Alkyl radicals
Methyl radical Ethyl radical n-Propyl radical Isopropyl radical
CH
3
C
2
H
5
C
3
H
7
C
3
H
7
H
|
|
C
H
H
|
H
|
|
C
H
H
|
|
|
C
H
H
|
H
|
|
C
H
H
|
|
|
C
H
H
|
|
|
C
H
H
|
|
H
H
|
|
C
H
H
|
|
C
H
|
|
|
C
H
H
|
Hydroxyalkyl radicals Acyl (alkanoyl) radicals
Hydroxyethyl radical Formyl radical Acetyl radical
(ethanyl radical) (methanoyl radical) (ethanoyl radical)
HOCH
2
CH
2
HCO CH
3
CO
O
H
|
|
C
H
H
|
|
C
H
|
|
H
|
O
C
H
O
H
|
|
C
H
H
|
|
C
Aryl radicals
Phenyl radical Methylphenyl radical Benzyl radical
C
6
H
5
C
6
H
4
(CH
3
) C
6
H
5
CH
2
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
CH
3
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
2
CH
|
|
Alkoxy radicals
Methoxy radical Ethoxy radical n-Propoxy radical
CH
3
O C
2
H
5
O C
3
H
7
O
H
|
| |
C O
H
H
|
H
|
|
|
C O
H
H
|
|
|
C
H
H
|
H
|
|
|
C
O
H
H
|
|
|
C
H
H
|
|
|
C
H
H
|
Hydroxyalkoxy radicals Acyloxy (alkanoyloxy) radicals
Hydroxyethyloxy radical Acetyloxy radical
(ethanoloxy radical) CH
3
C(O)O
HOCH
2
CH
2
O
O
H
|
| |
C O
H
H
|
|
|
C
H
H
|
O
|
H
|
|
C
O
H
H
|
|
C
(cont.)
723
Appendix B Tables
Table B.3 (cont.)
Chemical name
Molecular formula
Chemical structure
Organic
Aryloxy radicals
Phenoxy radical Methylphenoxy radical Benzoxy radical
PHO CRO C
6
H
5
CH
2
O
C
6
H
5
O 2-OC
6
H
4
CH
3
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
O
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
CH
3
O
|
|
|
|
C H
2
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
|
O
|
|
Alkylperoxy radicals
Methylperoxy radical Ethylperoxy radical n-Propylperoxy radical Alkylperoxy radical from
CH
3
O
2
C
2
H
5
O
2
C
3
H
7
O
2
OH addition to isoprene
ISOH
O
|
H
|
|
C
O
H
H
|
|
H
|
|
C
H
H
|
|
|
|
C
H
H
|
|
O
O
O
|
H
|
|
C
O
H
H
|
|
|
|
C
H
H
|
|
|
C
H
H
|
C
O
|
C
2
|
H
C
2
HO
H
|
CH
3
CH
|
|
O
|
Hydroxyalkylperoxy radicals
Hydroxymethylperoxy Hydroxyethylperoxy Hydroxypropylperoxy
radical radical radical
(methanolperoxy (ethanolperoxy (propanolperoxy
radical) radical) radical)
HOCH
2
O
2
HOC
2
H
4
O
2
HOC
3
H
6
O
2
O
OH
|
H
|
|
C
H
|
|
O
O
OH
|
H
|
|
C
H
|
|
O
|
|
C
H
H
|
O
OH
|
H
|
|
C
H
|
|
O
|
|
C
H
H
|
|
|
C
H
H
|
Acylperoxy (alkanoylperoxy) radicals
Peroxyacetyl radical Acetylmethylperoxy
CH
3
C(O)OO radical
CH
3
COCH
2
OO
O
O
| |
H
|
|
C
O
H
H
|
|
C
O
O
|
H
|
|
C
H
H
|
|
O
|
C
H
|
|
|
C
H
H
|
724
Table B.3 (cont.)
Chemical name
Molecular formula
Chemical structure
Organic
Arylperoxy radicals
Phenylperoxy radical Methylphenylperoxy Benzylperoxy radical Methylbenzylperoxy
PHO
2
radical BO
2
radical
C
6
H
5
OO CRO
2
C
6
H
5
CH
2
OO XLO
2
2-O
2
C
6
H
4
CH
3
2-CH
2
OOC
6
H
5
CH
3
O
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
O
|
|
C
O
H
3
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
O
|
|
|
|
|
O
C H
2
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
|
O
|
|
C
H
O
C
H
2
3
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
|
O
|
|
|
|
Other peroxy radicals
Peroxybenzoyl radical MVK+O
3
product 2-Methylperoxy MACR+O
3
product
BZO
2
OOCCH
3
CHCH
2
propenoyl radical CH
2
CCH
3
CHOO
C
6
H
5
OCOO CH
2
CCH
3
C(O)OO
|
O C H
2
|
CH
3
|
|
O
C
H
C
C
C
|
C O
H
3
2
H
|
|
O
|
H
C
O
O C
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
|
O
|
|
|
C
|
O
O
H
3
C
H
|
|
|
|
H
|
O
C C
Peroxy radical of MACR+OH product Nitrated organic peroxy
methylglyoxal CH
2
CCH
3
C(O)OO radical
MGPX PNO
2
CH
3
COC(O)OO
C O
O
|
O
H
3
C
2
CH
|
|
C
N
+
O
O
|
|
O
O
|
|
|
O
|
C C
O O
O
|
C H
3
|
|
C
O
|
C
OH adducts
Methylvinylketone-OH Methylvinylketone-OH Methacrolein-OH Methacrolein-OH
adduct adduct adduct adduct
MV1 MV2 MAC1 MAC2
OOCH
2
CH(OH) HOCH
2
CH(OO) OCHC(O
2
)(CH
3
) OCHC(OH)(CH
3
)
C(CH
3
)O C(CH
3
)O CH
2
OH CH
2
OO
C
O
O
|
O
H
3
|
H
C
2
|
H
|
|
|
|
O
C
H
C
C
O
H
|
O
H
3
|
H
C
2
|
O
|
|
|
|
O
C
H
C
C
O
O
|
O
H
3
|
H
C
2
|
|
|
|
H
|
O
C
H
C
C
O
O
|
O
H
3
|
H
C
2
|
|
|
|
H
|
O
C
H
C
(cont.)
725
Appendix B Tables
Table B.3 (cont.)
Chemical name
Molecular formula
Chemical structure
Organic
Toluene-OH adduct o-Xylene-OH adduct
TO
2
XINT
2-OHC
6
H
5
(O
2
)CH
3
4-OH-3-O
2
-2-
C
|
O
O
OH
H
3
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
H
|
CH
3
C
6
H
4
CH
3
C
|
H
O O
H
3
HO
CH
3 |
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
H
|
|
|
|
| | |
|
|
|
|
|
Molozonides Biradicals
Ethene molozonide Propene molozonide Criegee biradical Methyl criegee biradical
OZD OZD H
2
COO CH
3
HCOO
CH
2
O
3
CH
2
CH
3
CHO
3
CH
2
|
|
O
|
O
|
H
H
C |
|
O
|
C
O
H
3
|
H
C
O
O
C H
2 2
CH
| |
|
|
O
|
O
O
| CH
2
CH
C H
3
| |
|
|
O
|
Dicarbonyls Organic peroxides
Glyoxal (ethanedial) Methylglyoxal Methyl hydroperoxide
(CHO)
2
MGLY CH
3
OOH
CH
3
COCHO
O
C
H
|
| C
O
|
H
O
C
|
| C
O
|
H C H
3
|
|
O H
O H
|
|
|
C
H
H
|
Nitrates
Methyl nitrate Ethyl nitrate Propyl nitrate Benzyl nitrate
CH
3
ONO
2
C
2
H
5
ONO
2
C
3
H
7
ONO
2
C
6
H
5
CH
2
ONO
2
+
|
O
O
|
N O H
|
|
|
C
H
H
|
|
O
O
|
N O H
|
| |
C
H
H
|
|
|
C
H
H
|
+
|
O
O
|
N O H
|
| |
C
H
H
|
|
|
C
H
H
|
|
|
C
H
H
|
+
|
O
O
O C H
2 |
|
|
|
|
|
|
|
| |
|
|
|
|
|
N
|
|
|
|
|
|
|
Peroxyacetyl nitrate 2-Methylperoxy Peroxybenzoyl nitrate Isoprene-NO
3
adduct
PAN propenoylnitrate PBZN ISNT
CH
3
C(O)OONO
2
MPAN C
6
H
5
C(O)OONO
2
C
5
H
8
ONO
2
CH
2
CCH
3
C(O)OONO
2
C
|
|
O
H
|
|
C
H
H
|
|
+
O
O
|
N
O
O
|
C
|
C
|
H
3
|
|
|
O
C C
H
H
|
+
O
O
|
N
O
O
|
+
|
O
O
O
O
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
N
|
C O
|
|
|
C
C
|
H
3
C H
2
2
|
|
|
C
H
|
C
H
+
O
O
|
N O
726
Table B.3 (cont.)
Chemical name
Molecular formula
Chemical structure
Organic
Nitric acids
Methylperoxy nitric acid Ethylperoxy nitric acid
CH
3
O
2
NO
2
C
2
H
5
O
2
NO
2
| |
O
+
O
O
|
N
O
H
|
|
|
C
H
H
|
| |
O
+
O
O
|
N
O
H
|
|
|
C
H
H
|
|
|
C
H
H
|
Nitrites
Methyl nitrite Ethyl nitrite Propyl nitrite Benzyl nitrite
CH
3
ONO C
2
H
5
ONO C
3
H
7
ONO C
6
H
5
CH
2
ONO
O
O
N H
|
| |
C
H
H
|
|
O
O
N H
|
| |
C
H
H
|
|
|
|
C
H
H
|
O
O
N
H
|
| |
C
H
H
|
|
|
|
C
H
H
|
|
|
C
H
H
|
N
|
C H
2
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
|
O
O
|
|
Nitro group
Nitrobenzene m-Nitrotoluene o-Nitrophenol m-Nitrocresol
C
6
H
5
NO
2
3-NO
2
C
6
H
4
CH
3
NPHN NCRE
2-NO
2
C
6
H
4
OH 3-NO
2
-2-OHC
6
H
3
CH
3
+
O
O
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
N
|
|
+
O
O
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
CH
3
N
|
|
|
+
O
O
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
N
OH
|
|
|
|
+
O
O
|
|
|
|
|
|
|
|
|
|
|
|
| |
|
|
|
|
|
CH
3
N
OH
|
|
|
|
|
|
Carbon-bond groups
Parafn carbon bond Olen carbon bond Ketone carbonyl group Organic peroxide
PAR OLE KET ROOH
C
C C
O
C
|
|
H
|
O
O R
Primary organic peroxy Secondary organic oxy Secondary organic C
2
dinitrate group
radical radical peroxy radical DNIT
RO
2
ROR RO
2
R
N |
+
|
O
O
|
N
O
O
|
|
C
|
+
O
O
|
C

|
|
O
O R
O H
|
|
C
R
R
|
|
|
O
O
H
|
|
C
R
R
|
|
Organic nitrate
NTR
+
O
O
|
N
|
O
|
R
(cont.)
727
Appendix B Tables
Table B.3 (cont.)
Chemical name
Molecular formula
Chemical structure
Organic
Miscellaneous species and groups
Dimethyl 2nd organic Peroxide radical High-molecular- Dinitrate of isoprene
peroxide radical of OPEN weight aromatic oxidation DISN
AO
2
OPPX ring fragments
OPEN
Group of hydroxy Organic nitrates from Organic nitrates from Alkyl peroxy radical
carbonyl alkenes from OH addition to OH addition to from OH addition
isoprene-OH reaction unsubstituted substituted double bond across the double
IALD1 double bond and OH reactions of bond of ISNI1
ISNI1 primary products and ISNI2
ISNI2 ISNIR
Aromatic ring fragment Organic peroxide from Parafn to peroxy Parafn loss operator
acid isoprene radical operator XOP
ACID IPRX DOP
Molecular formulae and abbreviated names (boldface) used in Appendix Table B.4 are dened here.
Organic sulfur-, chlorine-, and bromine-containing species are included in the inorganic categories
for convenience.
B.4 GAS-PHASE REACTIONS
Table B.4 Gas-phase chemical kinetic reactions, reaction rates,
and photoprocesses
Rate coefcient
No. Region Kinetic reaction F
a
c
(s
1
, cm
3
s
1
, or cm
6
s
1
) Ref.
b
Inorganic chemistry
1 All O+O
2
+MO
3
+M 6.0010
34
(300/T )
2.3
B
2 S O+O
3
2O
2
8.0010
12
e
2060T
A
3 S O(
1
D) +O
3
2O
2
1.2010
10
B
4 S O(
1
D) +O
3
O
2
+2O 1.2010
10
B
5 All O(
1
D) +O
2
O+O
2
3.2010
11
e
67T
A
6 All O(
1
D) +N
2
O+N
2
1.8010
11
e
107T
A
7 S O(
1
D) +N
2
+MN
2
O+M 3.5010
37
(300T)
0.6
B
8 S O(
1
D) +N
2
ON
2
+O
2
4.4010
11
A
9 S O(
1
D) +N
2
ONO+NO 7.2010
11
A
10 All O(
1
D) +H
2
OH+H 1.1010
10
A
11 All O(
1
D) +H
2
OOH+OH 2.2010
10
A
12 All H+O
2
M
HO
2
(P) 0.55 5.4010
32
(300T)
1.8
A
7.5010
11
728
B.4 Gas-phase reactions
Table B.4 (cont.)
Rate coefcient
a
c
(s
1
, cm
3
s
1
, or cm
6
s
1
) Ref.
b
Inorganic chemistry
13 All H+O
3
O
2
+OH 1.40 10
10
e
470T
B
14 All H+HO
2
H
2
+O
2
5.6010
12
A
15 All H+HO
2
OH+OH 7.2010
11
A
16 All H+HO
2
H
2
O+O 2.4010
12
A
17 All OH+OH+O
2
2.3010
11
e
110T
A
18 All OH+O
3
HO
2
+O
2
1.9010
12
e
1000T
A
19 All OH+H
2
H
2
O+H 7.7010
12
e
2100T
A
20 All OH+OHH
2
O+O 4.2010
12
e
240T
B
21 All OH+OH
M
H
2
O
2
(P) 0.5 6.90 10
31
(300T)
0.8
A
2.610
11
22 All OH+HO
2
H
2
O+O
2
4.8010
11
e
250T
A
23 All OH+H
2
O
2
HO
2
+H
2
O 2.9010
12
e
160T
A
24 All OH+NO
M
HONO (P) 0.90 7.40 10
31
(300T)
2.4
A
4.5010
11
25 All OH+NO
2
M
HNO
3
(P) 0.43 2.60 10
30
(300T)
2.9
A
6.7010
11
(300T)
0.6
26 All OH+NO
3
HO
2
+NO
2
2.0010
11
A
27 All OH+HONOH
2
O+NO
2
1.8010
11
e
390T
A
28 All OH+HNO
3
H
2
O+NO
3
c
A
29 All OH+HO
2
NO
2
H
2
O+NO
2
+O
2
1.5010
12
e
360T
A
30 All OH+COHO
2
+CO
2
d
A
31 All HO
2
+OOH+O
2
2.7010
11
e
224T
A
32 All HO
2
+O
3
OH+2O
2
1.4010
14
e
600T
A
33 All HO
2
+HO
2
H
2
O
2
+O
2
e
F
34 All HO
2
+NOOH+NO
2
3.7010
12
e
240T
A
35 All HO
2
+NO
2
M
HO
2
NO
2
(P) 0.60 1.80 10
31
(300T)
3.2
A
4.7010
12
(300T)
1.4
36 All HO
2
+NO
3
HNO
3
+O
2
4.0010
12
A
37 All H
2
O
2
+OOH+HO
2
1.4010
12
e
2000T
A
38 All NO+O
M
NO
2
(P) 0.85 1.00 10
31
(300T)
1.6
A
3.0010
11
(300T)
0.3
39 All NO+O
3
NO
2
+O
2
1.8010
12
e
1370T
A
40 All NO
2
+ONO+O
2
6.5010
12
e
120T
A
41 All NO
2
+O
M
NO
3
(P) 0.80 9.00 10
32
(300T)
2.0
A
2.2010
11
42 All NO
2
+O
3
NO
3
+O
2
1.2010
13
e
2450T
A
43 All NO
3
+ONO
2
+O
2
1.7010
11
A
44 All NO
3
+NO2NO
2
1.8010
11
e
110T
A
45 All NO
3
+NO
2
M
N
2
O
5
(P) 0.33 2.70 10
30
(300T)
3.4
A
2.0010
12
(300T)
0.2
46 All N
2
O
5
M
NO
3
+NO
2
(P) 0.33 1.00 10
3
(300T)
3.5
A
e
11000T
9.7010
14
(300T)
0.1
e
11080T
47 All N
2
O
5
+H
2
O2HNO
3
2.0010
21
A
48 All HO
2
NO
2
M
HO
2
+NO
2
(P) 0.60 5.00 10
6
e
10000T
A
2.6010
15
e
10900T
(cont.)
729
Appendix B Tables
Table B.4 (cont.)
Rate coefcient
No. Region Kinetic Reaction F
a
c
(s
1
, cm
3
s
1
, or cm
6
s
1
) Ref.
b
Organic Chemistry
Alkane, alkene, and aldehyde chemistry
49 S CH
4
+O(
1
D) CH
3
O
2
+OH 1.4010
10
A
50 S CH
4
+O(
1
D) HCHO+H
2
1.5010
11
A
51 All CH
4
+OHCH
3
O
2
+H
2
O 2.3010
12
e
1765T
A
52 All CH
3
O+O
2
HCHO+HO
2
7.2010
14
e
1080T
A
53 U/T CH
3
O+NOHCHO+HO
2
+NO 4.0010
12
(300T)
0.7
A
54 U/T CH
3
O+NO
M
CH
3
ONO (P) 0.60 1.6010
29
(300T)
3.5
A
3.6010
11
(300T)
0.6
55 U/T CH
3
O+NO
2
M
CH
3
ONO
2
(P) 0.44 2.8010
29
(300T)
4.5
A
2.0010
11
56 U/T CH
3
ONO
2
+OHHCHO 1.0010
14
e
1060T
A
+NO
2
+H
2
O
57 All CH
3
O
2
+HO
2
CH
3
OOH+O
2
3.8010
13
e
780T
A
58 All CH
3
O
2
+NOCH
3
O+NO
2
4.2010
12
e
180T
A
59 U/T CH
3
O
2
+NO
2
M
CH
3
O
2
NO
2
(P) 0.36 2.5010
30
(300T)
5.5
A
7.5010
12
60 U/T CH
3
O
2
NO
2
M
CH
3
O
2
+NO
2
(P) 0.36 9.0010
5
e
9690T
A
1.1010
16
e
10560T
61 U/T CH
3
O
2
+CH
3
O
2
2CH
3
O+O
2
5.9010
13
e
509T
A
62 U/T CH
3
O
2
+CH
3
O
2
HCHO+CH
3
OH 7.0410
14
e
365T
A
63 U/T CH
3
O
2
+CH
3
C(O)OOCH
3
O
2
5.1010
12
e
272T
A
+CH
3
O+CO
2
64 All CH
3
OOH+OHCH
3
O
2
+H
2
O 1.9010
12
e
190T
A
65 U/T C
2
H
6
+OHC
2
H
5
O
2
+H
2
O 7.9010
12
e
1030T
A
66 U/T C
2
H
5
O
2
+NOC
2
H
5
O+NO
2
8.7010
12
A
67 U/T C
2
H
5
O
2
+NO
2
M
C
2
H
5
O
2
NO (P) 0.31 1.3010
29
(300T)
6.2
A
8.8010
12
68 U/T C
2
H
5
O
2
NO
2
M
C
2
H
5
O
2
+NO
2
(P) 0.31 4.8010
4
e
9285T
A
8.8010
15
e
10440T
69 U/T C
2
H
5
O
2
+HO
2
ROOH+O
2
2.7010
13
e
1000T
A
70 U/T C
2
H
5
O+O
2
CH
3
CHO+HO
2
6.0010
14
e
550T
A
71 U/T C
2
H
5
O+NOC
2
H
5
ONO 4.4010
11
A
72 U/T C
2
H
5
O+NOCH
3
CHO+HO
2
+NO 1.3010
11
A
73 U/T C
2
H
5
O+NO
2
C
2
H
5
ONO
2
2.8010
11
A
74 U/T C
3
H
8
+OHC
3
H
7
O
2
+H
2
O 8.0010
12
e
590T
A
75 U/T C
3
H
7
O
2
+NOC
3
H
7
O+NO
2
4.8010
12
A
76 U/T C
3
H
7
O+O
2
CH
3
COCH
3
+HO
2
1.5010
14
e
200T
A
77 U/T C
3
H
7
O+NOC
3
H
7
ONO 3.4010
11
A
78 U/T C
3
H
7
O+NOCH
3
COCH
3
+HO
2
+NO 6.5010
12
A
79 U/T C
3
H
7
O+NO
2
C
3
H
7
ONO
2
3.5010
11
A
80 U/T C
2
H
4
+OH
M
HOC
2
H
4
O
2
(P) 0.70 7.0010
29
(300/T)
3.1
A
9.0010
12
81 U/T HOC
2
H
4
O
2
+NONO
2
6.9310
12
C
+2HCHO+H
82 U/T HOC
2
H
4
O
2
+NONO
2
2.0710
12
C
+CH
3
CHO+OH
83 U/T C
2
H
4
+O
3
HCHO+H
2
COO 3.4010
15
e
2580T
A
84 U/T C
2
H
4
+O
3
HCHO+HCOOH
5.7010
15
e
2580T
A
85 U/T H
2
COO+NONO
2
+HCHO 7.0010
12
C
730
Table B.4 (cont.)
Rate coefcient
a
c
(s
1
, cm
3
s
1
, or cm
6
s
1
) Ref.
b
Organic chemistry
86 U/T H
2
COO+H
2
OHCOOH+H
2
O 4.0010
16
C
87 U/T H
2
COO+HCHOOZD 2.00 10
12
C
88 U/T H
2
COO+CH
3
CHOOZD 2.00 10
12
C
89 U/T HCOOH
CO
2
+H
2
0.21 A
90 U/T HCOOH
CO+H
2
O 0.60 A
91 U/T HCOOH
OH+HO
2
+CO 0.19 A
92 U/T C
3
H
6
+OH
M
HOC
3
H
6
O
2
(P) 0.50 8.00 10
27
(300T)
3.5
A
3.0010
11
93 U/T HOC
3
H
6
O
2
+NONO
2
+CH
3
CHO 6.00 10
12
C
+HCHO+HO
2
94 U/T C
3
H
6
+O
3
HCHO+CH
3
HCOO 4.10 10
16
e
1880T
A
95 U/T C
3
H
6
+O
3
HCHO+CH
3
HCOO
2.3410
15
e
1880T
A
96 U/T C
3
H
6
+O
3
CH
3
CHO+H
2
COO 1.03 10
15
e
1880T
A
97 U/T C
3
H
6
+O
3
CH
3
CHO+H
2
COO
1.7210
15
e
1880T
A
98 U/T CH
3
HCOO+NONO
2
+CH
3
CHO 7.00 10
12
C
99 U/T CH
3
HCOO+H
2
OCH
3
COOH+H
2
O 4.0010
16
C
100 U/T CH
3
HCOO+HCHOOZD 2.00 10
12
C
101 U/T CH
3
HCOO+CH
3
CHOOZD 2.00 10
12
C
102 U/T CH
3
COOH
CH
4
+CO
2
0.16 A
103 U/T CH
3
COOH
CH
3
O
2
+CO+OH 0.64 A
104 U/T CH
3
COOH
CH
3
O+CO+HO
2
0.20 A
105 All HCHO+OHHO
2
+CO+H
2
O 8.6010
12
e
20T
A
106 All HCHO+OOH+HO
2
+CO 3.4010
11
e
1600T
B
107 U/T HCHO+NO
3
HNO
3
+HO
2
+CO 5.8010
16
A
108 U/T HCHO+HO
2
HOCH
2
O
2
9.7010
15
e
625T
A
109 U/T HOCH
2
O
2
HO
2
+HCHO 2.40 10
12
e
7000T
A
110 U/T HOCH
2
O
2
+HO
2
ROOH 5.60 10
15
e
2300T
A
111 U/T HOCH
2
O
2
+NONO
2
+HO
2
7.0010
12
C
+HCOOH
112 U/T CH
3
CHO+OCH
3
C(O)OO+OH 1.80 10
11
e
1100T
B
113 U/T CH
3
CHO+OHCH
3
C(O)OO+H
2
O 5.6010
12
e
310T
A
114 U/T CH
3
CHO+NO
3
CH
3
C(O)OO+HNO
3
1.4010
12
e
1860T
A
115 U/T CH
3
C(O)OO+HO
2
ROOH+O
2
1.1410
13
e
1040T
A
116 U/T CH
3
C(O)OO+HO
2
CH
3
O
2
+OH 3.1610
13
e
1040T
A
+CO
2
117 U/T CH
3
C(O)OO+NONO
2
+CH
3
O
2
2.0010
11
A
+CO
2
118 U/T CH
3
C(O)OO+NO
2
M
CH
3
C(O)OONO
2
(P) 0.30 2.70 10
28
(300T)
7.1
A
1.2010
11
(300T)
0.9
119 U/T CH
3
C(O)OO+CH
3
C(O)OO2CH
3
O
2
2.8010
12
e
530T
A
+O
2
120 U/T CH
3
C(O)OONO
2
M
CH
3
C(O)OO+NO
2
(P) 0.30 4.90 10
3
e
12100T
A
5.4010
16
e
13830T
121 U/T CH
3
COCH
3
+OHCH
3
COCH
2
OO 2.8010
12
e
760T
A
+H
2
O
122 U/T CH
3
COCH
2
OO+NOCH
3
C(O)OO 8.10 10
12
C
+HCHO+NO
2
(cont. )
731
Appendix B Tables
Table B.4 (cont.)
Rate coefcient
a
c
(s
1
, cm
3
s
1
, or cm
6
s
1
) Ref.
b
Organic chemistry
123 U/T CH
3
OH+OHHCHO+HO
2
+H
2
O 2.6310
12
e
360T
A
124 U/T CH
3
OH+OHCH
3
O+H
2
O 4.6610
13
e
360T
A
125 U/T C
2
H
5
OH+OHCH
3
CHO+HO
2
+H
2
O 3.6910
12
e
70T
A
126 U/T PAR+OHRO
2
+H
2
O 9.2010
14
C
127 U/T PAR+OHRO
2
R+H
2
O 7.2010
13
C
128 U/T RO
2
+NONO
2
+HO
2
+CH
3
CHO 7.7010
12
C
+XOP
129 U/T RO
2
+NONTR 4.4010
11
e
1400T
C
130 U/T RO
2
R+NONO
2
+ROR 7.0010
12
C
131 U/T RO
2
R+NONTR 1.2010
10
e
1400T
C
132 U/T ROR+NO
2
NTR 1.5010
11
C
133 U/T RORKET+HO
2
1.6010
3
C
134 U/T RORKET+DOP 2.1010
14
e
8000T
C
135 U/T RORCH
3
CHO+DOP+XOP 4.0010
14
e
8000T
C
136 U/T RORCH
3
COCH
3
+DOP+2XOP 4.4010
14
e
8000T
C
137 U/T XOP+PAR 6.8010
12
C
138 U/T DOP+PARRO
2
5.1010
12
C
139 U/T DOP+PARAO
2
+2XOP 1.5010
12
C
140 U/T DOP+PARRO
2
R 1.7010
13
C
141 U/T DOP+KETCH
3
C(O)OO+XOP 6.8010
12
C
142 U/T AO
2
+NONO
2
+CH
3
COCH
3
+HO
2
8.1010
12
C
143 U/T OLE+O2PAR 4.1010
12
e
324T
C
144 U/T OLE+OCH
3
CHO 4.1010
12
e
324T
C
145 U/T OLE+OHO
2
+CO+RO
2
1.2010
12
e
324T
C
146 U/T OLE+ORO
2
+XOP+CO+HCHO+OH 2.4010
12
e
324T
C
147 U/T OLE+OHCH
3
O
2
+CH
3
CHO+XOP 5.2010
12
e
504T
C
148 U/T OLE+O
3
CH
3
CHO+H
2
COO+XOP 2.8010
15
e
2105T
C
149 U/T OLE+O
3
HCHO+CH
3
HCOO+XOP 2.8010
15
e
2105T
C
150 U/T OLE+O
3
CH
3
CHO+HCOOH
+XOP 4.3010
15
e
2105T
C
151 U/T OLE+O
3
HCHO+CH
3
COOH
+XOP 4.3010
15
e
2105T
C
152 U/T OLE+NO
3
PNO
2
7.7010
15
C
153 U/T PNO
2
+NODNIT 6.8010
13
C
154 U/T PNO
2
+NO HCHO+CH
3
CHO 6.8010
12
C
+XOP+2NO
2
Aromatic chemistry
155 U TOL+OHBO
2
+H
2
O 1.7010
13
e
322T
C
156 U TOL+OHCRES +HO
2
7.6010
13
e
322T
C
157 U TOL+OHTO
2
1.2010
12
e
322T
C
158 U BO
2
+ NONO
2
+BZA+HO
2
8.1010
12
C
159 U BZA+OHBZO
2
+H
2
O 1.3010
11
C
160 U BZO
2
+ NONO
2
+PHO
2
+ CO
2
2.5010
12
C
161 U BZO
2
+ NO
2
PBZN 8.4010
12
E
162 U PBZNBZO
2
+ NO
2
1.6010
15
e
13033T
E
163 U PHO
2
+ NONO
2
+PHO 8.1010
12
C
164 U PHO+NO
2
NPHN 1.3010
11
e
300T
E
165 U CRES +OHCRO+H
2
O 1.6010
11
C
166 U CRES +OHCRO
2
+H
2
O 2.5010
11
C
732
Table B.4 (cont.)
Rate coefcient
a
c
(s
1
, cm
3
s
1
, or cm
6
s
1
) Ref.
b
Organic chemistry
Aromatic chemistry
167 U CRES +NO
3
CRO+HNO
3
2.2010
11
C
168 U CRO+NO
2
NCRE 1.4010
11
C
169 U CRO
2
+NONO
2
+OPEN+HO
2
4.0010
12
C
170 U CRO
2
+ NONO
2
+ACID+HO
2
4.0010
12
C
171 U TO
2
+ NONO
2
+OPEN+HO
2
7.3010
12
C
172 U TO
2
+NONTR 8.1010
13
C
173 U TO
2
HO
2
+CRES 4.20 C
174 U XYL+OHCRES +PAR+HO
2
3.3210
12
e
116T
C
175 U XYL+OHXLO
2
+ H
2
O 1.7010
12
e
116T
C
176 U XYL+OHTO
2
5.0010
12
e
116T
C
177 U XYL+OHXINT 6.6010
12
e
116T
C
178 U XLO
2
+NONO
2
+HO
2
+BZA+PAR 8.1010
12
C
179 U XINT+NONO
2
+HO
2
+2CH
3
COCHO+PAR 8.1010
12
C
180 U CH
3
COCHO+OHMGPX+H
2
O 1.5010
11
A
181 U MGPX+NONO
2
+CH
3
C(O)OO+CO
2
8.1010
12
C
182 U OPEN+OHOPPX+CH
3
C(O)OO+HO
2
+CO 3.0010
11
C
183 U OPEN+O
3
CH
3
CHO+MGPX+HCHO+CO 1.6010
18
e
500T
C
184 U OPEN+O
3
HCHO+CO+OH+2HO
2
4.3010
18
e
500T
C
185 U OPEN+O
3
CH
3
COCHO 1.1010
17
e
500T
C
186 U OPEN+O
3
CH
3
C(O)OO+HCHO+HO
2
+CO 3.2010
17
e
500T
C
187 U OPEN+O
3
5.4010
18
e
500T
C
188 U OPPX+NONO
2
+HCHO+HO
2
+CO 8.1010
12
C
Terpene chemistry
189 U/T ISOP +OHISOH 2.5510
11
e
410T
D
190 U/T ISOP +O
3
0.67 MACR+0.26 MVK+0.3 OH 1.2310
14
e
2013T
D
+0.07 PAR+0.07 OLE+0.07 H
2
COO
+0.8 HCHO+0.06 HO
2
+0.15 CO
2
+ 0.05 CO
191 U/T ISOP +O0.22 MACR+0.63 MVK+0.08 ISOH 3.3010
11
D
192 U/T ISOP +NO
3
ISNT 7.8010
13
D
193 U/T ISOH+NO0.364 MACR+0.477 MVK 1.2210
11
e
180T
D
+0.840 HCHO
+0.08 ISNI1 +0.08 ISNI2
+0.886 HO
2
+0.840 NO
2
194 U/T ISNT+NO1.1 NO
2
+0.8 HO
2
+0.80 ISNI1 1.3910
11
e
180T
D
+0.1 MACR+0.15 HCHO
+0.05 MVK+0.05 DISN
195 U/T ISNI1 +OHISNIR 3.3510
11
D
196 U/T ISNI2 +OHISNIR 1.8810
11
D
197 U/T ISNIR+NO0.05 DISN+0.05 HO
2
+1.9 NO
2
1.3910
11
e
180T
D
+0.95 CH
3
CHO+0.95 CH
3
COCH
3
198 U/T ISNI1 +O
3
0.2 O+0.08 OH+0.5 HCHO 5.0010
18
D
+0.5 IALD1 +0.5 ISNI2 +0.5 NO
2
199 U/T ISOH+ISOH0.6 MACR+0.6 MVK+1.2 HCHO 2.0010
13
D
+1.2 HO
2
200 U/T ISOH+HO
2
IPRX 6.1510
11
e
900T
D
(cont.)
733
Appendix B Tables
Table B.4 (cont.)
Rate coefcient
a
c
(s
1
, cm
3
s
1
, or cm
6
s
1
) Ref.
b
Organic chemistry
Terpene chemistry
201 U/T IPRX+OHISOH 2.0010
11
D
202 U/T IPRX+O
3
0.7 HCHO 8.0010
18
D
203 U/T MACR+O
3
0.8 CH
3
COCHO+0.7 HCHO 5.3210
15
e
2520T
D
+0.2 O +0.09 H
2
COO+0.2 CO
+0.275 HO
2
+0.215 OH
+0.16 CO
2
+0.15 CH
2
CCH
3
CHOO
204 U/T MVK+O
3
0.82 CH
3
15
e
2016T
D
+0.2 O+0.11 H
2
COO+0.05 CO
+0.06 HO
2
+0.08 OH
+0.04 CH
3
CHO
+0.07 OOCCH
3
CHCH
2
205 U/T MACR+OH0.42 MAC1+0.08 MAC2 1.86 10
11
e
175T
D
+0.5 CH
2
CCH
3
C(O)OO
206 U/T MVK+OH0.28 MV1+0.72 MV2 4.1110
12
e
453T
D
207 U/T MAC1+NO0.95 HO
2
1.3910
11
e
180T
D
+ 0.95 CO+0.95 CH
3
COCH
3
+0.95 NO
2
+0.05 ISNI2
208 U/T MAC2+NO0.95 HO
2
1.3910
11
e
180T
D
+0.95 HCHO
+0.95 CH
3
COCHO
+0.95 NO
2
+0.05 ISNI2
209 U/T MV1+NO0.95 CH
3
11
e
180T
D
+0.05 ISNI2
+0.95 NO
2
+0.95 HO
2
210 U/T MV2+NO0.95 CH
3
CHO 1.3910
11
e
180T
D
+0.95 CH
3
C(O)OO+0.05 ISNI2
+0.95 NO
2
211 U/T MV1+HO
2
ROOH 6.1510
11
e
900T
D
212 U/T MV2+HO
2
ROOH 6.1510
11
e
900T
D
213 U/T MAC1+HO
2
ROOH 6.1510
11
e
900T
D
214 U/T MAC2+HO
2
ROOH 6.1510
11
e
900T
D
215 U/T CH
2
CCH
3
C(O)OO+NO
2
MPAN 8.4010
12
D
216 U/T MPANCH
2
CCH
3
C(O)OO+NO
2
1.5810
16
e
13507T
D
217 U/T CH
2
CCH
3
C(O)OO+NOC
2
H
4
+CH
3
O
2
1.4010
11
D
+NO
2
+CO
2
Sulfur chemistry
218 All SO
2
+OH
M
HSO
3
(P) 0.45 4.00 10
31
(300T)
3.3
A
2.0010
12
219 All HSO
3
+O
2
SO
3
+HO
2
1.3010
12
e
330T
A
220 All SO
3
+H
2
OH
2
SO
4
6.0010
15
A
221 T CH
3
SCH
3
+OHCH
3
SCH
2
+H
2
O 1.1310
11
e
254T
A
222 T CH
3
SCH
3
+OHCH
3
S(OH)CH
3
f
A
223 T CH
3
SCH
2
+O
2
CH
3
SCH
2
O
2
7.3010
13
N
224 T CH
3
SCH
2
O
2
+NOCH
3
SCH
2
O+NO
2
8.0010
12
N
225 T CH
3
SCH
2
OCH
3
S +HCHO 1.0010
1
N
734
Table B.4 (cont.)
Rate coefcient
a
c
(s
1
, cm
3
s
1
, or cm
6
s
1
) Ref.
b
Sulfur chemistry
226 T CH
3
S +O
2
CH
3
SOO
2.510
18
A
227 T CH
3
SOO
+NOCH
3
SO+NO
2
1.410
11
N
228 T CH
3
SOO
CH
3
S +O
2
6.010
2
N
229 T CH
3
SOCH
3
O
2
+SO 5.010
5
N
230 T CH
3
SO+O
3
CH
3
SO
2
+O
2
2.010
12
N
231 T CH
3
SO
2
CH
3
O
2
+SO
2
1.110
1
N
232 T CH
3
S(OH)CH
3
CH
3
SOH+CH
3
O
2
5.010
5
N
233 T CH
3
S(OH)CH
3
+OHCH
3
S(O)
2
CH
3
5.810
11
N
+2HO
2
234 T CH
3
SOH+OHCH
3
SO+H
2
O 1.110
10
N
235 T CH
3
SSCH
3
+OHCH
3
SOH+CH
3
S 7.0010
11
e
350T
A
236 T CH
3
SH+OHCH
3
S +H
2
O 9.9010
12
e
356T
A
237 T H
2
S +OHHS +H
2
O 6.3010
12
e
80T
A
238 T HS +O
2
SO+OH 4.0010
19
A
239 T/S SO+O
2
SO
2
+ O 1.6010
13
e
2280T
A
240 T/S OCS +OHHS +CO
2
1.1010
13
e
1200T
A
241 T/S CS
2
+OHHS +OCS 2.0010
15
A
242 T/S CS +O
2
OCS +O 2.9010
19
A
243 T/S S +O
2
SO+O 2.1010
12
A
Chlorine chemistry
244 S Cl +O
2
+ N
2
ClOO+N
2
1.4010
33
(300T)
3.9
A
245 S Cl +O
3
ClO+O
2
2.9010
11
e
260T
A
246 S Cl +H
2
HCl +H 3.7010
11
e
2300T
A
247 S Cl +HO
2
HCl +O
2
1.8010
11
e
170T
A
248 S Cl +H
2
O
2
HCl +HO
2
1.1010
11
e
980T
A
249 S Cl +NO
2
M
ClNO
2
(P) 0.60 1.8010
31
(300T)
2.0
B
1.0010
10
(300T)
1.0
250 S Cl +HNO
3
HCl +NO
3
2.0010
16
A
251 S Cl +NO
3
ClO+NO
2
2.4010
11
A
252 S Cl +CH
4
HCl +CH
3
O
2
9.6010
12
e
1350T
A
253 S Cl +HCHOHCl +CO+HO
2
8.2010
11
e
34T
A
254 S Cl +HOCl Cl
2
+ OH 2.5010
12
e
130T
B
255 S Cl +OClOClO+ClO 3.4010
11
e
160T
A
256 S Cl +ClOOCl
2
+ O
2
2.3010
10
B
257 S Cl +ClOOClO+ClO 1.2010
11
B
258 S ClO+O Cl +O
2
3.8010
11
e
70T
A
259 S ClO+O
3
ClOO+O
2
1.5010
17
A
260 S ClO+O
3
OClO+O
2
1.0010
18
A
261 S ClO+OHCl +HO
2
9.9010
12
e
120T
B
262 S ClO+OHHCl +O
2
1.1010
12
e
120T
B
263 S ClO+HO
2
HOCl +O
2
4.6610
13
e
710T
B
264 S ClO+HO
2
HCl +O
3
1.4410
14
e
710T
B
265 S ClO+NOCl +NO
2
6.2010
12
e
294T
A
266 S ClO+NO
2
M
ClONO
2
(P) 0.50 1.6010
31
(300T)
3.4
A
2.0010
11
(300T)
1.9
267 S ClO+ClOCl +ClOO 3.0010
11
e
2450T
A
(cont.)
735
Appendix B Tables
Table B.4 (cont.)
Rate coefcient
a
c
(s
1
, cm
3
s
1
, or cm
6
s
1
) Ref.
b
Chlorine chemistry
268 S ClO+ClO
M
Cl
2
O
2
(P) 0.60 1.7010
32
(300T)
4.0
A
5.4010
12
269 S HCl +OHCl +H
2
O 2.4010
12
e
330T
A
270 S ClONO
2
+OCl +NO
2
3.0010
12
e
800T
A
271 S OClO+OClO+O
2
2.4010
12
e
960T
A
272 S OClO+OHHOCl +O
2
4.5010
13
e
800T
A
273 S OClO+NOClO+NO
2
2.5010
12
e
600T
B
274 S ClOO+MCl +O
2
+M 2.8010
10
e
1820T
A
275 S HOCl +OClO+OH 1.0010
11
e
1300T
A
276 S HOCl +OHClO+H
2
O 3.0010
12
e
500T
A
277 S Cl
2
O
2
M
ClO+ClO (P) 0.60 1.0010
6
e
8000T
A
4.8010
15
e
8820T
Bromine chemistry
278 S Br +O
3
BrO+O
2
1.7010
11
e
800T
A
279 S Br +HO
2
HBr +O
2
1.4010
11
e
590T
A
280 S Br +H
2
O
2
HBr +HO
2
1.0010
11
e
3000T
B
281 S Br +HCHOHBr +CO+HO
2
1.7010
11
e
800T
A
282 S BrO+OBr +O
2
1.9010
11
e
230T
A
283 S BrO+OHBr +HO
2
7.510
11
B
284 S BrO+HO
2
HOBr +O
2
3.4010
12
e
540T
B
285 S BrO+NOBr +NO
2
8.7010
12
e
260T
A
286 S BrO+NO
2
M
BrONO
2
(P) 0.4 4.7010
31
(300T)
3.1
A
1.7010
11
(300T)
0.6
287 S BrO+ClOBr +OClO 1.6010
12
e
430T
A
288 S BrO+ClOBr +ClOO 2.9010
12
e
220T
A
289 S BrO+ClOBrCl +O
2
5.8010
13
e
170T
A
290 S BrO+BrO2Br +O
2
2.4010
12
e
40T
B
291 S BrO+BrOBr
2
+O
2
2.8010
14
e
860T
B
292 S BrO+O
3
Br +2O
2
1.0010
12
e
3200T
B
293 S HBr +OHBr +H
2
O 1.1010
11
(298T)
0.8
A
294 S HOBr +OBrO+OH 1.4010
10
e
430T
H
295 S BrCl +OBrO+Cl 2.2010
10
I
Photoprocesses
J (25 km) J (0 km)
Region Photolysis reaction (m) (s
1
) (s
1
) Ref.
b
296 S O
2
+h O+O -0.245 1.210
11
4.3610
26
B
297 All O
3
+h O(
1
D) +O
2
-0.31 1.13 10
4
5.0810
5
B
298 All O
3
+h O+O
2
>0.31 4.95 10
4
4.1710
4
B
299 S H
2
O+h H+OH -0.21 6.61 10
11
0 A
300 All H
2
O
2
+h 2OH -0.355 1.18 10
5
7.7210
6
A
301 All NO
2
+h NO+O -0.42 1.24 10
2
8.8210
3
B
302 All NO
3
+h NO
2
+O 0.410.67 3.04 10
1
2.8410
1
A
303 All NO
3
+h NO+O
2
0.590.63 2.65 10
2
2.4910
2
A
304 S N
2
O+h N
2
+O(
1
D) -0.24 2.49 10
8
1.8110
23
B
305 All N
2
O
5
+h NO
2
+NO
3
-0.385 4.73 10
5
5.0410
5
B
736
Table B.4 (cont.)
Photoprocesses
J (25 km) J (0 km)
No. Region Photolysis reaction (m) (s
1
) (s
1
) Ref.
b
306 All HONO+h OH+NO -0.40 2.7510
3
1.9410
3
A
307 All HNO
3
+h OH+NO
2
-0.335 5.6110
6
8.2310
7
A
308 All HO
2
NO
2
+h HO
2
+NO
2
-0.33 1.1410
5
3.6810
6
A
309 All HO
2
NO
2
+h OH+NO
3
-0.33 5.6710
6
1.8410
6
A
310 All HCHO+h 2HO
2
+CO -0.334 6.1710
5
3.2910
5
A
311 All HCHO+h CO+H
2
-0.37 7.6410
5
4.4010
5
A
312 All CH
3
OOH+h CH
3
O+OH -0.36 9.9210
6
5.7310
6
A
313 U/T CH
3
CHO+h CH
3
O
2
+HO
2
+CO -0.325 1.3010
5
6.3510
6
A
314 U/T CH
3
ONO
2
+h CH
3
O+NO
2
-0.33 2.5410
6
1.2510
6
A
315 U/T CH
3
O
2
NO
2
+h CH
3
O
2
+NO
2
-0.325 1.7610
5
6.4310
6
A
316 U/T C
2
H
5
ONO
2
+h C
2
H
5
O+NO
2
-0.330 1.4910
5
1.7910
6
A
317 U/T C
3
H
7
ONO
2
+h C
3
H
7
O+NO
2
-0.330 1.9410
5
3.4410
6
A
318 U/T CH
3
CO
3
NO
2
+h CH
3
CO
3
+NO
2
-0.300 3.5410
6
7.2310
8
A
319 U/T CH
3
COCH
3
+h CH
3
O
2
-0.335 1.9210
6
8.3910
7
A
+CH
3
C(O)OO
320 U/T KET+h CH
3
C(O)OO+RO
2
-0.33 1.9210
6
8.3910
7
J
+2XOP
321 U/T MVK+h CH
3
C(O)OO -0.325 8.1110
7
4.1210
7
K
+C
2
H
4
+HO
2
322 U/T MACR+h C
2
H
4
+HO
2
-0.375 2.5910
6
1.6410
6
A
+CO+CH
3
O
2
323 U/T CH
3
COCHO+h CH
3
C(O)OO -0.465 3.8210
4
2.9410
4
A
+CO+HO
2
324 U BZA+h PHO
2
+CO+HO
2
-0.385 6.4510
3
4.3110
5
E
325 U OPEN+h CH
3
C(O)OO -0.33 5.1310
4
2.9410
4
G
+CO+HO
2
326 T CH
3
SSCH
3
+h 2CH
3
S -0.40 1.3010
4
6.2810
5
A
327 S OCS +h CO+S -0.26 2.8310
8
1.7210
21
A
328 T/S CS
2
+h CS +S -0.34 2.4610
5
1.4610
5
A
329 S HCl +h H+Cl -0.22 2.1310
8
9.2510
24
A
330 S ClO+h Cl +O -0.305 1.3010
4
4.0610
5
L
331 S ClOO+h ClO+O -0.280 6.2210
8
2.8110
16
B
332 S OClO+h ClO+O -0.45 1.3610
1
9.8610
2
A
333 S HOCl +h OH+Cl -0.375 4.1010
4
2.5910
4
A
334 S ClONO
2
+h Cl +NO
3
-0.40 6.2210
5
4.7510
5
A
335 S Cl
2
+h Cl +Cl -0.45 3.5110
3
2.3710
3
A
336 S Cl
2
O
2
+h Cl +ClOO -0.36 2.0210
3
1.2110
3
A
337 S ClNO
2
+h Cl +NO
2
-0.372 6.5010
4
3.9410
4
A
338 S CH
3
Cl +h CH
3
O
2
+Cl -0.22 1.3110
8
2.2610
24
A
339 S CH
3
CCl
3
+h 3Cl -0.24 7.2010
7
5.1210
22
A
340 S CCl
4
+h Cl +CCl
3
-0.25 8.5110
7
2.1210
21
A
341 S CFCl
3
+h Cl +CFCl
2
-0.25 5.1010
7
4.3410
22
A
342 S CF
2
Cl
2
+h Cl +CF
2
Cl -0.226 4.6010
8
1.6510
23
A
343 S CFCl
2
CF
2
Cl +h Cl +CFCl
2
CF
2
-0.230 8.7810
8
2.6310
23
B
344 S CF
2
ClH+h Cl -0.205 5.6810
10
3.0410
27
B
345 S BrO+h Br +O -0.375 5.8810
3
3.8210
3
A
346 S HOBr +h Br +OH -0.48 9.0510
4
6.1210
4
A
347 S BrONO
2
+h Br +NO
3
-0.39 1.6110
3
1.0610
3
A
(cont.)
737
Appendix B Tables
Table B.4 (cont.)
Photoprocesses
J (25 km) J (0 km)
No. Region Photolysis reaction (m) (s
1
) (s
1
) Ref.
b
348 S Br
2
+h 2Br -0.60 3.5610
2
3.0210
2
B
349 S CH
3
Br +h CH
3
O
2
+Br -0.26 1.1310
6
3.9710
21
A
350 S BrCl +h Br +Cl -0.57 5.8810
3
4.6110
3
M
Species in the reactions are listed in Table B.3. Species above reaction arrows are second or third
bodies included in pressure- dependent reactions (footnote a). M= N
2
+O
2
is total air. The region
column indicates whether the reaction is important in urban areas (U), the free-troposphere (T),
the stratosphere (S), urban areas and the free-troposphere (U/T), or all three regions (All). The
Ref. column refers to sources of data for reaction rate coefcients, absorption cross sections, and
quantum yields. Photolysis coefcients at 0 and 25 km were calculated for a nonpolluted sky, a
zenith angle of 0
, and UV surface albedo of 0.03. The wavelengths given are valid for the listed
photoprocesses.
a
(P) indicates a pressure-dependent reaction, for which the reaction rate coefcient is
k
r
=
k
.T
k
0.T
[M]
k
.T
+k
0.T
[M]
F
1+
log
10
k
0.T
[M]
k
.T
2
1
c
where k
0.T
is the temperature-dependent three-body, low-pressure limit rate coefcient (the rst
rate listed), k
.T
is the two-body, high-pressure limit rate coefcient (the second rate listed),
[M] =[N
2
] + [O
2
] is the concentration (molecules cm
3
) of the third body, and F
c
is the broadening
factor.
b
A, Atkinson et al. (1997); B, DeMore et al. (1997); C, Gery et al. (1989); D, Paulson and Seinfeld
(1992); E, Carter (1991); F, Stockwell (1995); G, Gery et al. (1988); H, Nesbitt et al. (1995); I,
Clyne et al. (1976); J, assumed the same as for acetone; K, assumed the same as for methyl ethyl
ketone; L, Watson (1977); M, Seery and Britton (1964); N, Yin et al. (1990).
c
k
r
= k
1
+k
3
[M] (1 +k
3
[M] k
2
), where k
1
= 7.2010
15
e
785T
, k
2
= 4.1010
16
e
1440T
, k
3
=
1.9010
33
e
725T
, and [M] = [N
2
] +[O
2
] (molecules cm
3
).
d
k
r
= 1.3010
13
(1+0.6p
a.bar
)(300T)
1.0
, where p
a.bar
is the ambient air pressure in bars.
e
k
r
= (2.3010
13
e
600T
+1.7010
33
[M]e
1000T
)(1+1.4010
21
[H
2
O]e
2200T
), where [M]
= [N
2
] +[O
2
] and [H
2
O] are in units of molecules cm
3
.
f
k
r
= 1.710
42
[M]e
7810T
(1.0+5.510
31
[M] e
7460T
), where [M] = [N
2
] +[O
2
] (molecules
cm
3
).
B.5 CHEMICALS INVOLVED IN EQUILIBRIUM AND
AQUEOUS REACTIONS
Table B.5 List of species involved in equilibrium or aqueous-chemistry reactions
Chemical name Chemical formula Chemical name Chemical formula
Gas-phase precursors
Nitrogen dioxide NO
2
(g) Hydrochloric acid HCl(g)
Nitrate radical NO
3
(g) Carbon dioxide CO
2
(g)
Nitrous acid HONO(g) Formaldehyde HCHO(g)
Nitric acid HNO
3
(g) Formic acid HCOOH(g)
2
NO
2
(g) Methanol CH
3
OH(g)
Hydroxyl radical OH(g) Methylperoxy radical CH
3
O
2
(g)
738
B.5 Chemicals involved in equilibrium and aqueous reactions
Table B.5 (cont.)
Hydrogen peroxide H
2
O
2
(g) Methyl hydroperoxide CH
3
OOH(g)
Water vapor H
2
O(g) Acetic acid CH
3
COOH(g)
Ozone O
3
(g) Peroxyacetyl nitrate CH
3
C(O)OONO
2
(g)
Sulfur dioxide SO
2
(g) Peroxyacetic acid CH
3
C(O)OOH(g)
Sulfuric acid H
2
SO
4
(g) Methylglyoxal CH
3
COCHO(g)
Ammonia NH
3
(g) Nitrocresol C
6
H
3
(CH
3
)(OH)NO
2
(g)
Dissociated ionic molecules
Hydrogen ion H
+
Peroxysulfate radical ion SO
5
Ammonium ion NH
4
+
Peroxymonosulfate ion HSO
5
Sodium ion Na
+
Hydromethanesulfonate
ion
HOCH
2
SO
3
Potassium ion K
+
Oxymethanesulfonate ion

OCH
2
SO
3
Magnesium ion Mg
2+
Nitrate ion NO
3
Calcium ion Ca
2+
Nitrite ion NO
2
Hydroxy ion OH
Chloride ion Cl
Hydroperoxy ion HO
2
Dichloride ion Cl
2
Peroxy ion O
2
Chlorine hydroxide radical ClOH
Bisulfate ion HSO

4
Bicarbonate ion HCO

3
Sulfate ion SO
4
2
Carbonate ion CO
3
2
Bisulte ion HSO
3
Carbonate radical ion CO

3
Sulte ion SO
3
2
Formate HCOO
Sulfate radical ion SO

4
Acetate CH
3
COO
Undissociated molecules
Nitrogen dioxide NO
2
(aq) Carbon monoxide CO(aq)
Nitrate radical NO
3
(aq) Carbon dioxide CO
2
(aq)
Nitrous acid HONO(aq) Formaldehyde HCHO(aq)
Nitric acid HNO
3
(aq) Methylene glycol H
2
C(OH)
2
(aq)
2
NO
2
(aq) Formic acid HCOOH(aq)
Hydroxyl radical OH(aq) Methanol CH
3
OH(aq)
Hydroperoxy radical HO
2
(aq) Methylperoxy radical CH
3
O
2
(aq)
Hydrogen peroxide H
2
O
2
(aq) Methyl hydroperoxide CH
3
OOH(aq)
Water H
2
O(aq) Acetic acid CH
3
COOH(aq)
Molecular oxygen O
2
(aq) Peroxyacetyl nitrate CH
3
C(O)OONO
2
(aq)
Ozone O
3
(aq) Peroxyacetic acid CH
3
C(O)OOH(aq)
Sulfur dioxide SO
2
(aq) Methylglyoxal CH
3
COCHO(aq)
Sulfuric acid H
2
SO
4
(aq) Hydrated methylglyoxal CH
3
COCH
3
O
2
(aq)
Ammonia NH
3
(aq) Nitrocresol C
6
H
3
(CH
3
)(OH)NO
2
(aq)
Hydrochloric acid HCl(aq) Soluble manganese(II) ion
and complexes
Mn(II)
Chlorine atom Cl(aq) Soluble iron(III) ion and
complexes
Fe(III)
Solids
Ammonium nitrate NH
4
NO
3
(s) Potassium carbonate K
2
CO
3
(s)
Ammonium chloride NH
4
Cl(s) Magnesium nitrate Mg(NO
3
)
2
(s)
Ammonium bisulfate NH
4
HSO
4
(s) Magnesium chloride MgCl
2
(s)
Ammonium sulfate (NH
4
)
2
SO
4
(s) Magnesium sulfate MgSO
4
(s)
Triammonium bisulfate (NH
4
)
3
H(SO
4
)
2
(s) Magnesium carbonate MgCO
3
(s)
Ammonium bicarbonate NH
4
HCO
3
(s) Calcium nitrate Ca(NO
3
)
2
(s)
(cont.)
739
Appendix B Tables
Table B.5 (cont.)
Sodium nitrate NaNO
3
(s) Calcium chloride CaCl
2
(s)
Sodium chloride NaCl(s) Calcium sulfate CaSO
4
2H
2
O(s)
Sodium bisulfate NaHSO
4
(s) Calcium carbonate CaCO
3
(s)
Sodium sulfate Na
2
SO
4
(s) Manganese heptoxide Mn
2
O
7
(s)
Sodium bicarbonate NaHCO
3
(s) Silicon dioxide SiO
2
(s)
Sodium carbonate Na
2
CO
3
(s) Aluminum oxide Al
2
O
3
(s)
Potassium nitrate KNO
3
(s) Iron(III) oxide Fe
2
O
3
(s)
Potassium chloride KCl(s) Lead suboxide Pb
2
O(s)
Potassium bisulfate KHSO
4
(s) Elemental carbon EC(s)
Potassium sulfate K
2
SO
4
(s) Organic carbon OC(s)
Potassium bicarbonate KHCO
3
(s)
B.6 THERMODYNAMIC DATA
Table B.6 Values of L
f
G
o
i
, L
f
H
o
i
, and c
o
p.i
for some substances
Substance L
f
G
o
i
(kJ mol
1
) L
f
H
o
i
(kJ mol
1
) c
o
p.i
(J mol
1
K
1
)
NH
3
(g) 16.45
A
46.11
A
35.06
A
HNO
3
(g) 74.72
A
135.06
A
53.35
A
HCl(g) 95.299
A
92.307
A
29.126
A
SO
2
(g) 300.194
A
296.83
A
39.87
A
H
2
SO
4
(g) 690.289
B
814.21
B
CO
2
(g) 394.359
A
393.509
A
37.11
A
HCOOH(g) 351.0
A
378.57
A
CH
3
COOH(g) 374.0
A
432.25
A
66.5
A
H
2
O(aq) 237.129
A
285.83
A
75.291
A
NH
3
(aq) 26.5
A
80.29
A
35.06
E
HNO
3
(aq) 111.25
A
207.36
A
86.6
A
H
2
SO
3
(aq) 537.81
A
608.81
A
H
2
SO
4
(aq) 690.003
A
813.989
A
138.91
A
CO
2
(aq) 385.98
A
413.8
A
277.64
H
HCOOH(aq) 372.3
A
425.43
A
CH
3
COOH(aq) 396.46
A
485.76
A
H
+
0
A
0
A
0
A
NH
4
+
79.31
A
132.51
A
79.9
A
Na
+
261.905
A
240.12
A
46.4
A
K
+
283.27
A
252.38
A
21.8
A
Mg
2+
454.8
A
466.85
A
Ca
2+
553.58
A
542.83
A
OH
157.244
A
229.994
A
148.5
A
NO
3
111.25
F
207.36
F
86.6
F
Cl
131.228
A
167.159
A
136.4
A
HSO
4
755.91
A
887.34
A
84
A
SO
4
2
744.53
A
909.27
A
293.0
A
HSO
3
527.73
A
626.22
A
740
B.7 Equilibrium reactions and rate coefficients
Table B.6 (cont.)
Substance L
f
G
o
i
(kJ mol
1
) L
f
H
o
i
(kJ mol
1
) c
o
p.i
(J mol
1
K
1
)
SO
3
2
486.5
A
635.5
A
HCO
3
586.77
A
691.99
A
88.43
H
CO
3
2
527.81
A
677.14
A
234.52
H
HCOO
351.0
A
425.55
A
87.9
A
CH
3
COO
369.31
A
486.01
A
6.3
A
NH
4
NO
3
(s) 183.87
A
365.56
A
139.3
A
NH
4
Cl(s) 203.167
B
314.86
B
84.1
A
NH
4
HSO
4
(s) 823.0
C
1026.96
A
127.5
D
(NH
4
)
2
SO
4
(s) 901.67
A
1180.85
A
187.49
A
(NH
4
)
3
H(SO
4
)
2
(s) 1730.0
C
2207.0
C
315.0
D
NH
4
HCO
3
(s) 665.9
A
849.4
A
NaNO
3
(s) 367.0
A
467.85
A
92.88
A
NaCl(s) 384.24
B
411.26
B
50.5
A
NaHSO
4
(s) 1003.81
B
1132.19
B
85.0
D
Na
2
SO
4
(s) 1270.16
A
1387.08
A
128.2
A
NaHCO
3
(s) 851.0
A
950.81
A
87.61
A
Na
2
CO
3
(s) 1044.44
A
1130.68
A
112.3
A
KNO
3
(s) 394.86
A
494.63
A
96.4
A
KCl(s) 409.14
A
436.747
A
51.3
A
KHSO
4
(s) 1031.3
A
1160.6
A
87.16
G
K
2
SO
4
(s) 1321.37
A
1437.79
A
131.46
A
KHCO
3
(s) 863.5
A
963.2
A
89.27
H
K
2
CO
3
(s) 1063.5
A
1151.02
A
114.43
A
Mg(NO
3
)
2
(s) 589.4
A
790.65
A
141.92
A
MgCl
2
(s) 591.79
A
641.32
A
71.38
A
MgSO
4
(s) 1170.6
A
1284.9
A
96.48
A
MgCO
3
(s) 1012.1
A
1095.8
A
75.52
A
Ca(NO
3
)
2
(s) 743.07
A
938.39
A
149.37
A
CaCl
2
(s) 748.1
A
795.8
A
72.59
A
CaSO
4
2H
2
O(s) 1797.28
A
2022.63
A
186.02
A
CaCO
3
(s) 1128.79
A
1206.92
A
81.88
A
A, Wagman et al. (1982); B, Zaytsev and Aseyev (1992); C, Bassett and Seinfeld (1983); D, Wexler
and Seinfeld (1991); E, from A but for NH
2
+
; F, from A but for HNO
3
since NO
3
data may be
incorrect; G, Kim and Seinfeld (1995); H, Meng et al. (1995).
B.7 EQUILIBRIUM REACTIONS AND RATE COEFFICIENTS
Table B.7 Equilibrium reaction, rate coefcients,and coefcient units
No. Reaction A B C Units Ref.
1 SO
2
(g) SO
2
(aq) 1.22 10.55 mol kg
1
atm
1
A
2 H
2
O
2
(g) H
2
O
2
(aq) 7.45 10
4
22.21 mol kg
1
atm
1
B
3 O
3
(g) O
3
(aq) 1.13 10
2
7.72 mol kg
1
atm
1
C
4 NO
2
(g) NO
2
(aq) 1.00 10
2
8.38 mol kg
1
atm
1
D
(cont. )
741
Appendix B Tables
Table B.7 (cont.)
5 NO
3
(g) NO
3
(aq) 2.1010
5
29.19 mol kg
1
atm
1
E
6 OH(g) OH(aq) 2.5010
1
17.72 mol kg
1
atm
1
E
7 HO
2
(g) HO
2
(aq) 2.0010
3
22.28 mol kg
1
atm
1
E
8 HONO(g) HONO(aq) 4.9010
1
16.04 mol kg
1
atm
1
F
9 HO
2
NO
2
(g) HO
2
NO
2
(aq) 2.0010
4
mol kg
1
atm
1
G
10 HNO
3
(g) HNO
3
(aq) 2.1010
5
mol kg
1
atm
1
D
11 NH
3
(g) NH
3
(aq) 5.7610
1
13.79 5.39 mol kg
1
atm
1
A
12 HCHO(g) HCHO(aq) 3.46 8.19 mol kg
1
atm
1
H
13 HCOOH(g) HCOOH(aq) 5.3910
3
18.9 mol kg
1
atm
1
A
14 CO
2
(g) CO
2
(aq) 3.4110
2
8.19 28.93 mol kg
1
atm
1
A
15 CH
3
OH(g) CH
3
OH(aq) 2.2010
2
16.44 mol kg
1
atm
1
I
16 CH
3
O
2
(g) CH
3
O
2
(aq) 6.00 18.79 mol kg
1
atm
1
E
17 CH
3
OOH(g) CH
3
OOH(aq) 2.2710
2
18.82 mol kg
1
atm
1
B
18 CH
3
COOH(g) CH
3
COOH(aq) 8.60 10
3
21.58 mol kg
1
atm
1
A
19 CH
3
C(O)OOH(g) CH
3
C(O)OOH(aq) 4.73 10
2
20.70 mol kg
1
atm
1
B
20 CH
3
C(O)OONO
2
(g) CH
3
C(O)OONO
2
(aq) 2.90 19.83 mol kg
1
atm
1
J
21 CH
3
COCHO(g) CH
3
COCHO(aq) 3.70 10
3
25.33 mol kg
1
atm
1
K
22 HCHO(aq) H
2
C(OH)
2
(aq) 1.8210
3
13.49 atm
1
L
+H
2
O(aq)
23 SO
2
(aq) +H
2
O(aq) H
+
+HSO

3
1.7110
2
7.04 mol kg
1
A
24 CO
2
(aq) +H
2
O(aq) H
+
+HCO

3
4.3010
7
3.08 31.81 mol kg
1
A
25 Cl(aq) +H
2
O(aq) H
+
+ClOH
6.2010
8
mol kg
1
M
26 NH
3
(aq) +H
2
O(aq) NH
+
4
+OH
1.8110
5
1.50 26.92 mol kg
1
A
27 HNO
3
(aq) H
+
+NO

3
1.2010
1
29.17 16.83 mol kg
1
N
28 HCl(aq) H
+
+Cl
1.7210
6
23.15 mol kg
1
O
29 H
2
O(aq) H
+
+OH
1.0110
14
22.52 26.92 mol kg
1
A
30 H
2
SO
4
(aq) H
+
+HSO

4
1.0010
3
mol kg
1
R
31 H
2
O
2
(aq) H
+
+HO

2
2.2010
12
12.52 mol kg
1
S
32 HO
2
(aq) H
+
+O

2
3.5010
5
mol kg
1
R
33 HONO(aq) H
+
+NO

2
5.1010
4
4.23 mol kg
1
F
34 HCOOH(aq) H
+
+HCOO
1.8610
4
0.05 mol kg
1
A
35 CH
3
COOH(aq) H
+
+CH
3
COO
1.7510
5
0.10 mol kg
1
A
36 ClOH
Cl
+OH(aq) 1.43 mol kg

1
M
37 Cl

2
Cl(aq) +Cl
5.2610
6
mol kg
1
M
38 HSO

3
H
+
+SO
2
3
5.9910
8
3.74 mol kg
1
A
39 HSO

4
H
+
+SO
2
4
1.0210
2
8.85 25.14 mol kg
1
A
40 HCO

3
H
+
+CO
2
3
4.6810
11
5.99 38.84 mol kg
1
A
41 HNO
3
(g) H
+
+NO

3
2.5110
6
29.17 16.83 mol
2
kg
2
atm
1
A
42 HCl(g) H
+
+Cl
1.9710
6
30.19 19.91 mol
2
kg
2
atm
1
A
43 NH
3
(g) +H
+
NH
+
4
1.0310
11
34.81 5.39 atm
1
A
44 NH
3
(g) +HNO
3
(g) NH
+
4
+NO

3
2.5810
17
64.02 11.44 mol
2
kg
2
atm
2
A
45 NH
3
(g) +HCl(g) NH
+
4
+Cl
2.0310
17
65.05 14.51 mol
2
kg
2
atm
2
A
46 NH
4
NO
3
(s) NH
+
4
+NO

3
1.4910
1
10.40 17.56 mol
2
kg
2
A
47 NH
4
Cl(s) NH
+
4
+Cl
1.9610
1
6.13 16.92 mol
2
kg
2
A
48 NH
4
HSO
4
(s) NH
+
4
+HSO

4
1.3810
2
2.87 15.83 mol
2
kg
2
A
49 (NH
4
)
2
SO
4
(s) 2NH
+
4
+SO
2
4
1.82 2.65 38.57 mol
3
kg
3
A
50 (NH
4
)
3
H(SO
4
)
2
(s) 3NH
+
4
+HSO
4
+SO
2
4
2.9310
1
5.19 54.40 mol
5
kg
5
A
51 NH
4
HCO
3
(8) NH
+
4
+ HCO

3
1.08 10.04 mol
2
kg
2
A
52 NaNO
3
(s) Na
+
+NO

3
1.2010
1
8.22 16.01 mol
2
kg
2
A
53 NaCl(s) Na
+
+Cl
3.6110
1
1.61 16.90 mol
2
kg
2
A
742
B.8 Irreversible aqueous reactions
Table B.7 (cont.)
54 NaHSO
4
(s) Na
+
+HSO
4
2.8410
2
1.91 14.75 mol
2
kg
2
A
55 Na
2
SO
4
(s) 2Na
+
+SO
2
4
4.8010
1
0.98 39.50 mol
3
kg
3
A
56 NaHCO
3
(s) Na
+
+ HCO
3
3.9110
1
7.54 5.68 mol
2
kg
2
A
57 Na
2
CO
3
(s) 2Na
+
+ CO
2
3
1.8110
1
10.77 30.55 mol
3
kg
3
A
58 KNO
3
(s) K
+
+NO
3
8.7210
1
14.07 19.39 mol
2
kg
2
A
59 KCl(s) K
+
+Cl

8.68 6.94 19.95 mol
2
kg
2
A
60 KHSO
4
(s) K
+
+HSO
4
2.4010
1
8.42 17.96 mol
2
kg
2
A
61 K
2
SO
4
(s) 2K
+
+SO
2
4
1.5710
2
9.59 45.81 mol
3
kg
3
A
62 KHCO
3
(s) K
+
+HCO
3
1.4010
1
7.60 2.52 mol
2
kg
2
A
63 K
2
CO
3
(s) 2K
+
+CO
2
3
2.5410
5
12.46 36.73 mol
3
kg
3
A
64 Mg(NO
3
)
2
(s) Mg
2+
+ 2NO
3
2.51 10
15
mol
3
kg
3
A
65 MgCl
2
Mg
2+
+ 2Cl

9.55 10
21
mol
3
kg
3
A
66 MgSO
4
(s) Mg
2+
+SO
2
4
1.0810
5
mol
2
kg
2
A
67 MgCO
3
(s) Mg
2+
+ CO
2
3
6.8210
6
mol
2
kg
2
A
68 Ca(NO
3
)
2
(s) Ca
2+
+2NO
3
6.0710
5
mol
3
kg
3
A
69 CaCl
2
Ca
2+
+ 2Cl

7.9710
11
mol
3
kg
3
A
70 CaSO
4
2H
2
O(s) Ca
2+
+SO
2
4
4.3210
5
mol
2
kg
2
A
+2H
2
O(aq)
71 CaCO
3
(s) Ca
2+
+ CO
2
3
4.9710
9
mol
2
kg
2
A
The equilibrium coefcient reads,
K
eq
(T) = Aexp
T
0
T
1
+C
1
T
0
T
+ln
T
0
T
where
A= K
eq
(T
0
). B =
1
R
T
0
i
k
i
i
L
f
H
o
i
. C =
1
R
i
k
i
i
c
o
p.i
. and T
0
= 298.15 K
The terms in A, B, and C are dened in Chapter 17.
A, Derived from Appendix Table B.6 using (17.37); B, Lind and Kok (1986); C, Kozac-Channing
and Heltz (1983); D, Schwartz (1984); E, Jacob (1986); F, Schwartz and White (1981); G, Park
and Lee (1987); H, Ledbury and Blair (1925); I, Snider and Dawson (1985); J, Lee (1984); K,
Betterton and Hoffmann (1988); L, Le Henaff (1968); M, Jayson et al. (1973); N, derived from
a combination of other rate coefcients in the table; O, Marsh and McElroy (1985); R, Perrin
(1982); S, Smith and Martell (1976).
B.8 IRREVERSIBLE AQUEOUS REACTIONS
Table B.8 Aqueous-phase chemical kinetic reactions, rate coefcients, and photoprocesses
No. Reaction A B Ref.
1 S(IV) + H
2
O
2
(aq) + H
+
S(VI) + 2H
+
+ H
2
O(aq)
HSO
3
+ H
2
O
2
(aq) + H
+
7.45 10
7
(1) 15.96 A
(cont.)
743
Appendix B Tables
Table B.8 (cont.)
2 S(IV) + HO
2, T
S(VI) + OH(aq) + H
+
HSO
3
+ HO
2
(aq) 4.35 10
5
A
SO
3
2
+ HO
2
(aq) 5.65 10
5
A
HSO
3
+ O
2
4.35 10
4
A
SO
3
2
+ O
2
5.65 10
4
A
3 S(IV) + OH(aq) SO
5
+ H
2
O(aq)
HSO
3
+ OH(aq) 4.20 10
9
5.03 E
SO
3
2
+ OH(aq) 4.60 10
9
5.03 E
4 S(IV) + O
3
(aq) S(VI) + O
2
(aq) + H
+
SO
2
(aq) + O
3
(aq) 2.40 10
4
A
HSO
3
+ O
3
(aq) 3.70 10
5
18.56 A
SO
3
2
+ O
3
(aq) 1.50 10
9
17.72 A
5 S(IV) + HCHO
T
HOCH
2
SO
3
HSO
3
+ HCHO(aq) 7.90 10
2
16.44 G
SO
3
2
+ HCHO(aq) 2.48 10
7
6.04 G
6 S(IV) + Cl
2
SO
5
+ 2Cl
+ H
+
HSO
3
+ Cl
2
3.40 10
8
5.03 D
SO
3
2
+ Cl
2
1.60 10
8
5.03 D
7 S(IV) + SO
5
HSO
5
+ SO
5
HSO
3
+ SO
5
3.00 10
5
10.40 D
SO
3
2
+ SO
5
1.30 10
7
6.71 D
8 S(IV) + SO
4
S(VI) + SO
5
+ H
+
HSO
3
+ SO
4
1.30 10
9
5.03 H
SO
3
2
+ SO
4
5.30 10
8
5.03 H
9 S(IV) + HSO
5
2S(VI) + 2H
+
HSO
3
+ HSO
5
7.10 10
6
10.47 D
10 S(IV) + CH
3
OOH(aq) + H
+
S(VI) + 2H
+
+ CH
3
OH(aq)
HSO
3
+ CH
3
OOH(aq)+ H
+
1.90 10
7
12.75 A
11 S(IV) + CH
3
C(O)OOH(aq) S(VI) + H
+
+ CH
3
COOH
T
HSO
3
+ CH
3
C(O)OOH(aq) 3.60 10
7
(2) 13.42 E
12 S(IV) + Fe(III) S(VI) + H
2
O
2
(aq) + Fe(III)
SO
3
2
+ Fe(III) 9.50 10
7
20.27 B, C
13 S(IV) + Mn(II) S(VI) + H
2
O
2
(aq) + Mn(II) + H
+
HSO
3
+ Mn(II) 1.00 10
3
30.06 Cc
14 H
2
O
2
(aq) + OH(aq) H
2
O(aq) + HO
2, T
2.70 10
7
5.70 J
15 H
2
O
2
(aq) + Cl
2
2Cl
+ HO
2, T
+ H
+
1.40 10
5
11.31 Q
16 H
2
O
2
(aq) + Cl(aq) Cl
+ HO
2, T
+ H
+
4.50 10
7
R
17 H
2
O
2
(aq) + SO
4
HO
2, T
+ S(VI) + H
+
1.20 10
3
6.71 P
18 H
2
O
2
(aq) + CO
3
HO
2, T
+ CO
2, T
8.00 10
5
9.46 O
19 OH(aq) + HO
2, T
H
2
O(aq) + O
2
(aq)
OH(aq) + HO
2
(aq) 7.00 10
9
5.03 I
OH(aq) + O
2
1.00 10
10
5.03 I
20 OH(aq) + HSO
5
SO
5
+ H
2
O(aq) 1.70 10
7
6.38 H
21 OH(aq) + CO
2, T
H
2
O(aq) + CO
3
OH(aq) + HCO
3
1.50 10
7
6.41 M
744
B.8 Irreversible aqueous reactions
Table B.8 (cont.)
22 OH(aq) + HCHO
T
HO
2, T
+ HCOOH
T
+ H
2
O(aq)
OH(aq) + H
2
C(OH)
2
(aq) 2.00 10
9
5.03 U
23 OH(aq) + HOCH
2
SO
3
SO
5
+ HCHO
T
+ H
2
O(aq)
1.40 10
9
5.03 H
24 OH(aq) + HCOOH
T
HO
2, T
+ CO
2, T
+ H
2
O(aq)
OH(aq) + HCOOH(aq) 1.60 10
8
5.03 V
OH(aq) + HCOO
2.50 10
9
5.03 W
25 OH(aq) + CH
3
COOH
T
HO
2, T
+ CO (aq)
+ HCOOH
T
+ H
2
O(aq)
OH(aq) + CH
3
COOH(aq) 2.00 10
7
6.25 Y
OH(aq) + CH
3
COO
8.00 10
7
5.07 Y
26 OH(aq) + CH
3
OH(aq) HO
2, T
+ HCHO
T
+ H
2
O(aq)
4.50 10
8
5.03 W
27 OH(aq) + CH
3
OOH(aq) CH
3
O
2
(aq) 2.70 10
7
5.70 H
+ H
2
O(aq)
28 OH(aq) + CH
3
OOH(aq) HO
2, T
+ HCOOH
T
+ H
2
O(aq)
1.90 10
7
6.04 H
29 HO
2, T
+ O
3
(aq) OH(aq) + 2O
2
(aq)
O
2
+ O
3
(aq) 1.50 10
9
5.03 I
30 HO
2, T
+ HO
2, T
H
2
O
2
(aq) + O
2
(aq)
HO
2
(aq) + HO
2
(aq) 8.60 10
5
7.94 K
HO
2
(aq) + O
2
1.00 10
8
5.03 K
31 HO
2, T
+ CH
3
O
2
(aq) CH
3
OOH(aq) + O
2
(aq)
HO
2
(aq) + CH
3
O
2
(aq) 4.30 10
5
10.07 H
O
2
+ CH
3
O
2
(aq) 5.00 10
7
5.37 H
32 HO
2, T
+ SO
4
S(VI) + H
+
+ O
2
(aq)
HO
2
(aq) + SO
4
5.00 10
9
5.03 H
O
2
+ SO
4
5.00 10
9
5.03 H
33 HO
2, T
+ SO
5
HSO
5
+ O
2
(aq)
O
2
+ SO
5
1.00 10
8
5.03 H
34 HO
2, T
+ Cl
2
2Cl
+ O
2
(aq)
O
2
+ Cl
2
1.00 10
9
5.03 P
35 HO
2, T
+ CO
2, T
HO
2
+ CO
3
O
2
+ HCO
3
1.50 10
6
N
36 HO
2, T
+ CO
3
CO
2, T
+ O
2
(aq) + OH
O
2
+ CO
3
4.00 10
8
5.03 O
37 SO
4
+ HCOOH
T
S(VI) + CO
2, T
+ HO
2, T
SO
4
+ HCOO
1.10 10
8
5.03 AA
38 SO
4
+ Cl
S(VI) + Cl(aq) 2.60 10

8
5.03 AA
39 SO
5
+ HCOOH
T
HSO
5
+ CO
2, T
+ HO
2, T
SO
5
+ HCOO
1.40 10
4
13.42 H
40 SO
5
+ SO
5
2SO
4
+ O
2
(aq) 6.00 10
8
5.03 D
41 HCOOH
T
+ CO
3
2CO
2, T
+ HO
2, T
+ OH
HCOO
+ CO
3
1.10 10
5
11.41 X
(cont.)
745
Appendix B Tables
Table B.8 (cont.)
42 HCOOH
T
+ Cl
2
CO
2, T
+ HO
2, T
+ 2 Cl
HCOO
+ Cl
2
1.90 10
6
8.72 Q
43 H
2
O
2
(aq) + h 2OH(aq) Radiation dependent BB
44 NO
3
+ h NO
2
(aq) + OH(aq) + OH
Radiation dependent BB
Rate coefcients for individual reactions in each family reaction have the form
K
aq
(T) = Aexp
T
0
T
1
A=K
aq
(T
0
) B =
1
R
T
0
i
k
i
i
L
f
H
o
i
where A has units of s
1
, M
1
s
1
, or M
2
s
1
, the terms in B are dened in Chapter 17, T
0
=
298.15 K, and T is in kelvin. Specialized rate coefcients have the form
K
aq
(T) =
Aexp{B[(T
0
T) 1]}
1 +13[H
+
]
(1)
K
aq
(T) = Aexp
T
0
T
1
[H
+
] +7.0 10
2
(2)
where K
a1
= 0.0123 mol L
1
, K
a2
= 6.61 10
8
mol L
1
, and [H
+
] is in moles per liter. Rate
coefcients for family reactions are described in (19.56)(19.61).
A, Hoffmann and Calvert (1985); B, Conklin and Hoffmann (1988); C, Martin and Hill (1987a);
Cc, Martin and Hill (1987b); D, Huie and Neta (1987); E, Lind et al. (1987); G, Boyce and
Hoffmann (1984); H, Jacob (1986); I, Sehested et al. (1968); J, Christensen et al. (1982); K,
Bielski (1978); M, Weeks and Rabani (1966); N, Schmidt (1972); O, Behar et al. (1970); P, Ross
and Neta (1979); Q, Hagesawa and Neta (1978); R, Graedel and Goldberg (1983); U, Bothe and
Schulte-Frohlinde (1980); V, Scholes and Willson (1967); W, Anbar and Neta (1967); X, Chen
et al. (1973); Y, Farhataziz and Ross (1977); AA, Wine et al. (1989); BB, Graedel and Weschler
(1981).
B.9 SOLUTE ACTIVITY COEFFICIENT DATA
Table B.9 Parameters for calculating electrolyte mean binary solute activity coefcients
Electrolyte j B
j
G
j
H
j
HCl Hamer and Wu (1972) Parker (1965) Parker (1965)
16 m 55.5 m 15.9 m
0 1.998104 10
2
0 0
1 7.959068 101 5.532198 10
1
2.108728 10
0
2 6.580198 10
1
2.571126 10
1
8.542292 10
1
3 7.409941 10
2
2.790048 10
1
6.237459 10
1
4 1.345075 10
2
4.691631 10
2
1.935911 10
1
5 2.248651 10
3
2.382485 10
3
2.037543 10
2
HNO
3
Hamer and Wu (1972) Parker (1965) Parker (1965)
28 m 55.5 m 55.5 m
0 2.388378 10
2
0 0
1 7.777787 10
1
5.785894 10
1
4.785171 10
1
746
B.9 Solute activity coefficient data
Table B.9 (cont.)
Electrolyte j B
j
G
j
H
j
2 5.950086 10
1
9.860271 10
1
6.521896 10
0
3 1.284278 10
1
6.043012 10
1
2.605544 10
0
4 1.291734 10
2
1.123169 10
1
3.739984 10
1
5 6.257155 10
4
6.688134 10
3
1.832646 10
2
NaCl Hamer and Wu (1972) Parker (1965) Perron et al. (1981)
6.1 m 6.1 m 6.0 m
0 6.089937 10
3
0 0
1 1.015184 10
0
5.808744 10
1
2.261834 10
0
2 9.345503 10
1
1.163239 10
0
3.622494 10
0
3 4.615793 10
1
5.136893 10
1
1.608598 10
0
4 1.431557 10
1
1.029523 10
1
2.092972 10
1
5 1.700298 10
2
1.401488 10
2
0
NaNO
3
Wu and Hamer (1980) Parker (1965) Parker (1965)
10.8 m 9.2 m 2.2 m
0 6.638145 10
3
0 0
1 1.024329 10
0
5.678220 10
1
7.232987 10
1
2 6.877457 10
1
2.136826 10
0
1.918907 10
1
3 3.336161 10
1
1.145031 10
0
2.382164 10
1
4 8.387414 10
2
2.585350 10
1
1.367081 10
1
5 8.154844 10
3
2.390815 10
2
3.064556 10
0
NaHSO
4
Harvie et al. (1984) Assumed same as for NaCl Same as for NaCl
6.0 m 6.1 m 6.0 m
0 8.890979 10
3
0 0
1 9.559487 10
1
5.808744 10
1
2.261834 10
0
2 8.758970 10
1
1.163239 10
0
3.622494 10
0
3 4.607380 10
1
5.136893 10
1
1.608598 10
0
4 1.309144 10
1
1.029523 10
1
2.092972 10
1
5 1.398546 10
2
1.401488 10
2
0
Na
2
SO
4
Goldberg (1981) Wagman et al. (1982) Holmes and Mesmer
4.4 m 3.1 m (1986) 2.0 m
0 2.323071 10
2
0 0
1 3.321509 10
0
1.698182 10
0
9.410224 10
0
2 3.388793 10
0
5.160108 10
0
2.213823 10
1
3 2.402946 10
0
2.132810 10
0
3.481895 10
1
4 8.926764 10
1
8.840108 10
1
2.348397 10
1
5 1.225933 10
1
5.143058 10
1
6.471345 10
0
NH
4
Cl Hamer and Wu (1972) Wagman et al. (1982) Parker (1965)
7.4 m 7.0 m 7.4 m
0 5.022484 10
3
0 0
1 1.037873 10
0
4.890513 10
1
1.959107 10
0
2 8.517483 10
1
7.013315 10
1
9.894682 10
1
3 4.225323 10
1
4.682151 10
1
1.024499 10
1
4 1.214996 10
1
1.702461 10
1
2.354376 10
1
5 1.471525 10
2
2.502643 10
2
6.600384 10
2
(cont.)
747
Appendix B Tables
Table B.9 (cont.)
Electrolyte j B
j
G
j
H
j
NH
4
NO
3
Hamer and Wu (1972) Vanderzee et al. (1980) Roux et al. (1978)
25.9 m 25.0 m 22.4 m
0 1.044572 10
2
0 0
1 1.004940 10
0
4.362921 10
1
2.611682 10
0
2 4.674064 10
1
1.455444 10
0
3.158677 10
0
3 1.750495 10
1
6.282104 10
1
2.005748 10
0
4 3.253844 10
2
1.123507 10
1
4.113737 10
1
5 2.276789 10
3
7.438990 10
3
2.820677 10
2
NH
4
HSO
4
Bassett and Seinfeld Same as for NH
4
Cl Same as for NH
4
Cl
(1983) 6.0 m 7.0 m 7.4 m
0 2.708121 10
3
0 0
1 1.095646 10
0
4.890513 10
1
1.959107 10
0
2 1.042878 10
0
7.013315 10
1
9.894682 10
1
3 6.289405 10
1
4.682151 10
1
1.024499 10
1
4 2.079631 10
1
1.702461 10
1
2.354376 10
1
5 2.776957 10
2
2.502643 10
2
6.600384 10
2
(NH
4
)
2
SO
4
Filippov et al. (1985) Wagman et al. (1982) Sukhatme and
5.8 m 5.5 m Saikhedkar (1969)
5.5 m
0 2.163694 10
2
2.297972 10
1
0
1 3.377941 10
0
4.255129 10
1
1.609902 10
3
2 3.118007 10
0
2.220594 10
0
4.437758 10
0
3 1.920544 10
0
2.607601 10
0
6.101756 10
3
4 6.372975 10
1
1.243384 10
0
4.021805 10
1
5 8.277292 10
2
2.102563 10
1
4.375833 10
4
Parameters t into (17.56), which ts into (17.55). B, which also ts into (17.51), is used to calculate
binary solute activity coefcients at 298.15 K, G is a heat-of-dilution parameter, and H is a heat-
capacity parameter. Molalities are the maximum molalities of the data used to derive the ts.
From Jacobson et al. (1996b).
B.10 WATER ACTIVITY DATA
Table B.10 Parameters for calculating molalities of binary electrolytes as a function of
relative humidity at 298.15 K
A
HCl
A
HNO
3
B
H
+
/HSO
4
B
2H
+
/SO
4
2
Parameter 0% r.h.; 18.5 m 0% r.h.; 22.6 m 0% r.h.; 30.4 m 0% r.h.; 30.4 m
Y
0
1.874637647 10
1
2.306844303 10
1
3.0391387536 10
1
3.0391387536 10
1
Y
1
2.052465972 10
1
3.563608869 10
1
1.8995058929 10
2
1.8995058929 10
2
Y
2
9.485082073 10
1
6.210577919 10
1
9.7428231047 10
2
9.7428231047 10
2
Y
3
5.362930715 10
2
5.510176187 10
2
3.1680155761 10
3
3.1680155761 10
3
Y
4
1.223331346 10
3
1.460055286 10
3
6.1400925314 10
3
6.1400925314 10
3
Y
5
1.427089861 10
3
1.894467542 10
3
6.9116348199 10
3
6.9116348199 10
3
Y
6
8.344219112 10
2
1.220611402 10
3
4.1631475226 10
3
4.1631475226 10
3
Y
7
1.90992437 10
2
3.098597737 10
2
1.0383424491 10
3
1.0383424491 10
3
748
B.11 Surface resistance data
Table B.10 (cont.)
C
NaCl
C
NaNO
3
D
NaHSO
4
C
Na
2
SO
4
Parameter 47% r.h.; 13.5 m 30% r.h.; 56.8 m 1.9% r.h.; 158 m 58% r.h.; 13.1 m
Y
0
5.875248 10
1
3.1021762 10
2
1.8457001681 10
2
5.5983158 10
2
Y
1
1.8781997 10
2
1.82975944 10
3
1.6147765817 10
3
2.56942664 10
3
Y
2
2.7211377 10
2
5.13445395 10
3
8.444076586 10
3
4.47450201 10
3
Y
3
1.8458287 10
2
8.01200018 10
3
2.6813441936 10
4
3.45021842 10
3
Y
4
4.153689 10
1
7.07630664 10
3
5.0821277356 10
4
9.8527913 10
2
Y
5
3.33365806 10
3
5.5964847603 10
4
Y
6
6.5442029 10
2
3.2945298603 10
4
Y
7
8.002609678 10
3
A
NH
4
Cl
A
NH
4
NO
3
D
NH
4
HSO
4
C
(NH
4
)
2
SO
4
Parameter 47% r.h.; 23.2 m 62% r.h.; 28 m 6.5% r.h.; 165 m 37% r.h.; 29.0 m
Y
0
7.110541604 10
3
3.983916445 10
3
2.9997156464 10
2
1.1065495 10
2
Y
1
7.217772665 10
4
1.153123266 10
4
2.8936374637 10
3
3.6759197 10
2
Y
2
3.071054075 10
5
2.13956707 10
5
1.4959985537 10
4
5.0462934 10
2
Y
3
7.144764216 10
5
7.926990533 10
5
4.5185935292 10
4
3.1543839 10
2
Y
4
9.840230371 10
5
1.407853405 10
6
8.110895603 10
4
6.770824 10
1
Y
5
8.03407288 10
5
1.351250086 10
6
8.4994863218 10
4
Y
6
3.603924022 10
5
6.770046794 10
5
4.7928255412 10
4
Y
7
6.856992393 10
4
1.393507324 10
5
1.1223105556 10
4
Y coefcients t into (17.66). The table also lists the lowest relative humidity (r.h.) and corresponding
highest molality (m) for which each t is valid.
A, derived from Jacobson (1996b) and references therein; B, derived from Robinson and Stokes (1955);
C, obtained from Tang (1997); D, derived from Tang and Munkelwitz (1994).
B.11 SURFACE RESISTANCE DATA
Table B.11 Parameters used in surface resistance equations of Chapter 20
Parameter
(s m
1
) 1 2 3 4 5 6 7 8 9 10 11
Seasonal Category 1: Midsummer with lush vegetation
R
min
60 120 70 130 100 80 100 150
R
cut,0,q
2000 2000 2000 2000 2000 2500 2000 4000
R
canp
100 200 100 2000 2000 2000 0 0 300 150 200
R
soil.SO
2
400 150 350 500 500 100 0 1000 0 220 400
R
soil.O
3
300 150 200 200 200 300 2000 400 1000 180 200
R
surf.SO
2
2000 2000 2000 2000 2000 2500 2000 4000
R
surf.O
3
1000 1000 1000 1000 1000 1000 1000 1000
Seasonal Category 2: Autumn with unharvested cropland
R
min
250 500
R
cut,0,q
9000 9000 9000 4000 8000 9000 9000 9000
R
canp
100 150 100 1500 2000 1700 0 0 200 120 140
R
soil.SO
2
400 200 350 500 500 100 0 1000 0 300 400
R
soil.O
3
300 150 200 200 200 300 2000 400 800 180 200
R
surf.SO
2
9000 9000 9000 2000 4000 9000 9000 9000
R
surf.O
3
400 400 400 1000 600 400 400 400
(cont.)
749
Appendix B Tables
Table B.11 (cont.)
Parameter
(s m
1
) 1 2 3 4 5 6 7 8 9 10 11
Seasonal Category 3: Late autumn after frost, no snow
R
min
250 500
R
cut,0,q
9000 9000 4000 8000 9000 9000 9000
R
canp
100 10 100 1000 2000 1500 0 0 100 50 120
R
soil.SO
2
400 150 350 500 500 200 0 1000 0 200 400
R
soil.O
3
300 150 200 200 200 300 2000 400 1000 180 200
R
surf.SO
2
9000 9000 3000 6000 9000 9000 9000
R
surf.O
3
1000 400 400 1000 600 800 600 600
Seasonal Category 4: Winter, snow on ground and subfreezing
R
min
400 800
R
cut,0,q
6000 9000 9000 9000 9000
R
canp
100 10 10 1000 2000 1500 0 0 50 10 50
R
soil,SO
2
100 100 100 100 100 100 0 1000 100 100 50
R
soil,O
3
600 3500 3500 3500 3500 3500 2000 400 3500 3500 3500
R
surf,SO
2
9000 200 400 9000 9000
R
surf,O
3
1000 1000 400 1500 600 800 1000 800
Seasonal Category 5: Transitional spring with partially green short annuals
R
min
120 240 140 250 190 160 200 300
R
cut,0,q
4000 4000 4000 2000 3000 4000 4000 8000
R
canp
100 50 80 1200 2000 1500 0 0 200 60 120
R
soil.SO
2
500 150 350 500 500 200 0 1000 0 250 400
R
soil.O
3
300 150 200 200 200 300 2000 400 1000 180 200
R
surf.SO
2
4000 4000 4000 2000 3000 4000 4000 8000
R
surf.O
3
1000 500 500 1500 700 600 800 800
Column headings are the following land-use types: 1, urban land; 2, agricultural land; 3, range
land; 4, deciduous forest; 5, coniferous forest; 6, mixed forest, including wetland; 7, water, both
salt and fresh; 8, barren land, mostly desert; 9, nonforested wetland; 10, mixed agricultural and
range land; 11, rocky open areas with low-growing shrubs.
Source: Wesely (1989).
750
B.12 More surface resistance data
B.12 MORE SURFACE RESISTANCE DATA
Table B.12 Parameters used in surface resistance calculations
Chemical formula Chemical name D
v
D
q
H
q
(M atm
1
) f
0.q
SO
2
Sulfur dioxide 1.9 1 10
5
0
O
3
Ozone 1.6 1 10
2
1
NO
2
Nitrogen dioxide 1.6 1 10
2
0.1
NO Nitric oxide 1.3 3 10
3
0
HNO
3
Nitric acid 1.9 1 10
14
0
H
2
O
2
Hydrogen peroxide 1.4 1 10
5
1
CH
3
CHO Acetaldehyde 1.6 1.5 10
1
0
HCHO Formaldehyde 1.3 6 10
3
0
CH
3
OOH Methyl hydroperoxide 1.6 2.4 10
2
0.1
CH
3
C(O)OOH Peroxyacetic acid 2.0 5.4 10
2
0.1
HCOOH Formic acid 1.6 4 10
6
0
NH
3
Ammonia 0.97 2 10
4
0
CH
3
C(O)OONO
2
Peroxyacetyl nitrate 2.6 3.6 10
0
0.1
HONO Nitrous acid 1.6 1 10
5
0.1
Source: Wesely (1989).
751
References
Abdella K. and McFarlane N. (1997) A new second-order turbulence closure scheme for the
planetary boundary layer. J. Atmos. Sci. 54, 185067.
ACEA(1999) ACEAProgramme on Emissions of Fine Particles fromPassenger Cars. Brussels,
ACEA.
Adamson A. W. (1990) Physical Chemistry of Surfaces, 5th edn. New York, John Wiley and
Sons, Inc.
Alam M. K. (1987) The effect of van der Waals and viscous forces on aerosol coagulation.
Aerosol Sci. Technol. 6, 4152.
Alfaro S. C. and Gomes L. (2001) Modeling mineral aerosol production by wind erosion:
Emission intensities and aerosol size distributions in source areas. J. Geophys. Res. 106,
1807584.
Allen P. and Wagner K. (1992) 1987 California Air Resources Board emissions inventory,
magnetic tapes ARA806, ARA807.
Al Nakshabandi G. and Konhke H. (1965) Thermal conductivity and diffusivity of soils as
related to moisture tension and other physical properties. Agric. Meteor. 2, 2719.
Ambartzumiam V. (1936) The effect of the absorption lines on the radiative equilibrium of
the outer layers of the stars. Publ. Obs. Astron. Univ. Leningrad 6, 718.
Anandakumar K. (1999) A study of the partition of net radiation into heat uxes on a dry
asphalt surface. Atmos. Environ. 33, 391118.
Anbar M. and Neta P. (1967) A compilation of specic bimolecular rate constants for the
reactions of hydrated electrons, hydrogen atoms, and hydroxyl radicals with inorganic
and organic compounds in aqueous solution. Int. J. Appl. Radiat. Isot. 18, 493523.
Andre J. C., De Moor G., Lacarrere P., and Du Vachat R. (1978) Modeling the 24-hour
evolution of the mean and turbulent structures of the planetary boundary layer. J. Atmos.
Sci. 35, 186183.
Andreae M. O. and Merlet P. (2001) Emission of trace gases and aerosols from biomass
burning. Global Biogeochemical Cycles 15, 95566.
Andreas E. L. (1992) Sea spray and the turbulent air-sea heat uxes. J. Geophys. Res. 97,
11,42941.
Andr en A. (1990) Evaluation of a turbulence closure scheme suitable for air pollution appli-
cations. J. Appl. Math. Phys. 29, 22439.
Andres R. J. and Kasgnoc A. D. (1998) A time-averaged inventory of subaerial volcanic sulfur
emissions. J. Geophys. Res. 103, 2525161.
Angell C. A., Guni M. O., and Sichina W. J. (1982) Heat capacity of water at extremes of
supercooling and superheating. J. Phys. Chem. 86, 9981002.
Anthes R. A. (1977) A cumulus parameterization scheme utilizing a one-dimensional cloud
model. Mon. Wea. Rev. 105, 27086.
Anthes A., Kuo Y.-H., Hsie E.-Y., Low-Nam S., and Bettge T. W. (1989) Estimation of skill
and uncertainty in regional numerical models. Q. J. Roy. Meteor. Soc. 115, 763806.
Apsley D. D. and Castro I. P. (1997) A limited-length-scale k- model for the neutral and
stably-stratied atmospheric boundary layer. Boundary-Layer Meteor. 83, 7598.
752
References
Arakawa A. (1984) Boundary conditions in limited-area models. Course notes, Department
of Atmospheric Sciences, University of California, Los Angeles.
(1997) Adjustment mechanisms in atmospheric models. J. Meteor. Soc. Japan 75, 15579.
Arakawa A. and Konor C. S. (1995) Vertical differencing of the primitive equations based
on the CharneyPhillips grid in hybrid p vertical coordinates. Mon. Wea. Rev. 124,
51128.
Arakawa A. and Lamb V. R. (1977) Computational design of the basic dynamical processes
of the UCLA general circulation model. Methods Comput. Phys. 17, 174265.
Arakawa A. and Schubert W. H. (1974) Interaction of a cumulus cloud ensemble with large
scale environment, Part I. J. Atmos. Sci. 31, 674701.
Arakawa A. and Suarez. M. J. (1983) Vertical differencing of the primitive equations in sigma
coordinates. Mon. Wea. Rev. 111, 3445.
Archer C. L. and Jacobson M. Z. (2003) Spatial and temporal distributions of U.S. winds
and wind power at 80 m derived from measurements. J. Geophys. Res. 108 (D9), 4289,
doi:10.1029/2002JD002076.
Arking A. A. and Grossman K. (1972) The inuence of line shape and band structure on
temperatures in planetary atmospheres. J. Atmos. Sci. 29, 93749.
Arstila H., Korhonen P., and Kulmala M. (1999) Ternary nucleation: Kinetics and application
to waterammoniahydrochloric acid system. J. Aerosol Sci. 30, 1318.
Artelt C., Schmid H.-J., and Peukert W. (2003) On the relevance of accounting for the evolution
of the fractal dimension in aerosol process simulations. J. Aerosol Sci. 34, 51134.
Arya S. P. (1988) Introduction to Micrometeorology. San Diego, Academic Press, 307pp.
Asphalt Roong Manufacturers Association (ARMA) (1999) Roong Basics, http://www.
asphaltroong.org/basics.html.
Atkinson R., Lloyd A. C., and Winges L. (1982) An updated chemical mechanism for
hydrocarbon/NO
x
/SO
2
photooxidations suitable for inclusion in atmospheric simulation
models. Atmos. Environ. 16, 134155.
Atkinson R., Baulch D. L., Cox R. A., et al. (1997) Evaluated kinetic, photochemical, and
heterogeneous data for atmospheric chemistry. Supplement V. J. Phys. Chem. Ref. Data
26, 5211011.
Austin J. (1991) On the explicit versus family solution of the fully diurnal photochemical
equations of the stratosphere. J. Geophys. Res. 96, 12,94174.
Avissar R. and Mahrer Y. (1988) Mapping frost-sensitive areas with a three-dimensional
local-scale numerical model. Part I: Physical and numerical aspects. J. Appl. Meteor. 27,
40013.
Bader G. and Deuhard P. (1983) A semi-implicit mid-point rule for stiff systems of ordinary
differential equations. Numer. Math. 41, 37398.
Bagnold R. A. (1941) The Physics of Blown Sand and Desert Dunes. New York, Methuen,
265pp.
Baldocchi D. D., Hicks B. B., and Camara P. (1987) A canopy stomatal resistance model for
gaseous deposition to vegetated surfaces. Atmos. Environ. 21, 91101.
Bannon P. R. (1966) On the anelastic approximation for a compressible atmosphere. J. Atmos.
Sci. 53, 361828.
Bassett M. E. and Seinfeld J. H. (1983) Atmospheric equilibrium model of sulfate and nitrate
aerosol. Atmos. Environ. 17, 223752.
(1984) Atmospheric equilibriummodel of sulfate and nitrate aerosol-II. Particle size analysis.
Atmos. Environ. 18, 116370.
Bates T. S., Kiene R. P., Wolfe G. V., et al. (1994) The cycling of sulfur in surface seawater of
the Northeast Pacic. J. Geophys. Res. 99, 783543.
Beard K. V. (1976) Terminal velocity and shape of cloud and precipitation drops aloft.
J. Atmos. Sci. 33, 85164.
Beard K. V. and Grover S. N. (1974) Numerical collision efciencies for small raindrops
colliding with micron size particles. J. Atmos. Sci. 31, 54350.
753
References
Beard K. V. and Ochs III H. T. (1984) Collection and coalescence efciencies for accretion.
J. Geophys. Res. 89, 71659.
Beard K. V. and Pruppacher H. R. (1971) Awind tunnel investigation of the rate of evaporation
of small water drops falling at terminal velocity in air. J. Atmos. Sci. 28, 145564.
Behar D., Czapski G., and Duchovny I. (1970) Carbonate radical in ash photolysis and pulse
radiolysis of aqueous carbonate solutions. J. Phys. Chem. 74, 220610.
Bermejo R. and Conde J. (2002) A conservative quasi-monotone semi-Lagrangian scheme.
Mon. Wea. Rev. 130, 42330.
Berresheim H., Wine P. H., and Davis D. D. (1995) Sulfur in the atmosphere. In Composition,
Chemistry, and Climate of the Atmosphere. H. B. Singh, ed., New York, Van Nostrand
Reinhold, 251307.
Betterton E. A. and Hoffmann M. R. (1988) Henrys law constants of some environmentally
important aldehydes. Environ. Sci. Technol. 22, 141518.
Betts A. K. (1986) A new convective adjustment scheme. Part I: Observational and theoretical
basis. Q. J. Roy. Meteor. Soc. 112, 67791.
Betts A. K. and Miller M. J. (1986) Anewconvective adjustment scheme. Part II: Single column
tests using GATE wave, BOMEX, ATEX, and arctic air-mass data sets. Q. J. Roy. Meteor.
Soc. 112, 693709.
Bhumralkar C. M. (1975) Numerical experiments on the computation of ground surface
temperature in an atmospheric general circulation model. J. Appl. Meteor. 14, 67
100.
Bielski B. H. J. (1978) Reevaluation of the spectral and kinetic properties of HO
2
and O
2
free radicals. Photochem. Photobiol. 28, 6459.

Bigg E. K. (1953) The formation of atmospheric ice crystals by the freezing of droplets. Q. J.
Roy. Meteor. Soc. 79, 51019.
Binkowski F. S. and Roselle S. J. (2003) Models-3 Community Multiscale Air Quality (CMAQ)
model aerosol component 1. Model description. J. Geophys. Res. 108 (D6), 4183,
doi:10.1029/2001JD001409.
Binkowski F. S. and Shankar U. (1995) The regional particulate matter model 1. Model descrip-
tion and preliminary results. J. Geophys. Res. 100, 26191209.
Blackadar A. K. (1976) Modeling the nocturnal boundary layer. Proceedings of the Third
Symposium on Atmospheric Turbulence, Diffusion and Air Quality, Boston, American
Meteorological Society, 469.
(1978) Modeling pollutant transfer during daytime convection. Proceedings of the Fourth
Symposium on Atmospheric Turbulence, Diffusion, and Air Quality, Reno, American
Meteorological Society, 4437.
Blumthaler M. and Ambach W. (1988) Solar UVB-albedo of various surfaces. Photochem.
Photobiol. 48, 858.
Boccippio D. J., Cummins K. L., Christian H. J., and Goodman S. J. (2001) Combined satellite-
and surface-based estimation of the intracloudcloud-to-ground lightning ratio over the
continental United States. Mon. Wea. Rev. 129, 10822.
Bohren C. F. (1986) Applicability of effective-medium theories to problems of scattering and
absorption by nonhomogeneous atmospheric particles. J. Atmos. Sci. 43, 46875.
Bohren C. F. and Huffman D. R. (1983) Absorption and Scattering of Light by Small Particles.
New York, John Wiley and Sons, 530pp.
Bojkov R. D. and Fioletov V. E. (1995) Estimating the global ozone characteristics during the
last 30 years. J. Geophys. Res. 100, 16,53751.
Bolsaitis P. and Elliott J. F. (1990) Thermodynamic activities and equilibrium partial pressures
for aqueous sulfuric acid solutions. J. Chem. Eng. Data 35, 6985.
Bolton D. (1980) The computation of equivalent potential temperature. Mon. Wea. Rev. 108,
104653.
Bond D. W., Steiger S., Zhang R., Tie X., and Orville R. E. (2002) The importance of NOx
production by lightning in the tropics. Atmos. Environ. 36, 150919.
754
References
Bond T. C., Streets D. G., Yarber K. F., Nelson S. M., Woo, J.-H., and Klimont, Z. (2004) A
technology-based global inventory of black and organic carbon emissions from combus-
tion. J. Geophys. Res., 109, (D1) 4203, doi: 10.1029/2003JD003697.
Bonsang B., Martin D., Lambert G., Kanakidou M., Le Roulley J. C., and Sennequier G.
(1991) Vertical distribution of nonmethane hydrocarbons in the remote marine boundary
layer. J. Geophys. Res. 96, 731324.
Bothe E. and Schulte-Frohlinde D. (1980) Reaction of dihydroxymethyl radical with molecular
oxygen in aqueous solution. Z. Naturforsch. B, Anorg. Chem. Org. Chem. 35, 10359.
Bott A. (1989) A positive denite advection scheme obtained by nonlinear renormalization of
the advective uxes. Mon. Wea. Rev. 117, 100615.
(2000) Aux method for the numerical solution of the stochastic collection equation: Exten-
sion to two-dimensional particle distributions. J. Atmos. Sci. 57, 28494.
Bott A. and Carmichael G. R. (1993) Multiphase chemistry in a microphysical radiation fog
model a numerical study. Atmos. Environ. 27A, 50322.
Boubel R. W., Fox D. L., Turner D. B., and Stern A. C. S. (1994) Fundamentals of Air Pollution.
San Diego, Academic Press, Inc.
Boyce S. D. and Hoffmann M. R. (1984) Kinetics and mechanism of the formation of hydrox-
ymethanesulfonic acid at low pH. J. Phys. Chem. 88, 47406.
Brewer P. G. (1997) Ocean chemistry of the fossil fuel CO
2
signal: The haline signal of business
as usual. Geophys. Res. Lett. 24, 13679.
Briere S. (1987) Energetics of daytime sea breeze circulation as determined from a two-
dimensional and third-order closure mode. J. Atmos. Sci. 44, 145574.
Brock J. R., Zehavi D., and Kuhn P. (1986) Condensation aerosol formations and growth in
a laminar coaxial jet: Experimental. J. Aerosol Sci. 17, 1122.
Bromley L. A. (1973) Thermodynamic properties of strong electrolytes in aqueous solutions.
AIChE J. 19, 31320.
Brownawell M. (2004) http://isa.dknet.dk/innova/gemarkus.htm.
Bruggeman D. A. G. (1935) Berechnung verschiedener physikalischer Konstanten von hetero-
genen Substanzen. I. Dielektrizit atskonstanten und Leitf ahigkeiten der Mischk orper aus
isotropen Substanzen. Ann. Phys. (Leipzig) 24, 63979.
Brutsaert W. (1991) Evaporation in the Atmosphere. Dordrecht, Kluwer Academic Publishers,
299pp.
Burtscher H. and Schmidt-Ott A. (1982) Enormous enhancement of van der Waals forces
between small silver particles. Phys. Rev. Lett. 48, 17347.
Businger J. A., Wyngaard J. C., Izumi Y., and Bradley E. F. (1971) Flux-prole relationships
in the atmospheric surface layer. J. Atmos. Sci. 28, 1819.
Butler J. N. (1982) Carbon Dioxide Equilibria and Their Applications. Reading MA, Addison-
Wesley Publishing Co., 259pp.
Calder K. L. (1949) Eddy diffusion and evaporation in ow over aerodynamically smooth
and rough surfaces: A treatment based on laboratory laws of turbulent ow with special
reference to conditions in the lower atmosphere. Q. J. Mech. Appl. Math. 2, 15376.
Caldeira K. and Wickett M. E. (2003) Anthropogenic carbon and ocean pH. Nature 425, 265.
California Air Resources Board (CARB) (1988) Method Used to Develop a Size-Segregated
Particulate Matter Inventory. Technical Support Division, Emission Inventory Branch,
California Air Resources Board, Sacramento, CA.
Campbell F. W. and Maffel L. (1974) Contrast and spatial frequency. Sci. Am. 231, 10614.
Capaldo K. P., Pilinis C., and Pandis S. N. (2000) A computationally efcient hybrid approach
for dynamic gas/aerosol transfer in air quality models. Atmos. Environ. 34, 361727.
Carmichael G. R., Peters L. K., and Kitada T. (1986) A second generation model for regional-
scale transport/chemistry/deposition. Atmos. Environ. 20, 17388.
Carpenter R. L., Droegemeier K. K., Woodward P. R., and Hane C. E. (1990) Application of
the piecewise parabolic method (PPM) to meteorological modeling. Mon. Wea. Rev. 118,
586612.
755
References
Carter W. P. L. (1990) Adetailed mechanismfor the gas-phase atmospheric reactions of organic
compounds. Atmos. Environ. 24A, 481518.
(1991) Development of Ozone Reactivity Scales for Volatile Organic Compounds. EPA-
600/3-91-050. U.S. Environmental Protection Agency, Research Triangle Park, NC.
(2000) Documentation of the SAPRC-99 Chemical Mechanism for VOC Reactivity Assess-
ment. Final Report to the California Air Resources Board Under Contracts 92329 and
95308, California Air Resources Board, Sacramento, CA.
Cass G. R. (1979) On the relationship between sulfate air quality and visibility with examples
in Los Angeles. Atmos. Environ. 13, 106984.
Castro T., Madronich S., Rivale S., Muhlia A., and Mar B. (2001) The inuence of aerosols
on photochemical smog in Mexico City. Atmos. Environ. 35, 176572.
Celia M. A. and Gray W. G. (1992) Numerical Methods for Differential Equations. Englewood
Cliffs, Prentice-Hall.
Chameides W. L. (1984) The photochemistry of a remote marine stratiformcloud. J. Geophys.
Res. 89, 473955.
Chameides W. L. and Stelson A. W. (1992) Aqueous-phase chemical processes in deliquescent
sea-salt aerosols: A mechanism that couples the atmospheric cycles of S and sea salt. J.
Geophys. Res. 97, 20, 56580.
Chang E., Nolan K., Said M., Chico T., Chan S., and Pang E. (1991) 1987 Emissions Inventory
for the South Coast Air Basin: Average Annual Day. South Coast Air Quality Management
District (SCAQMD), Los Angeles.
Chang, S., Brodzinsky G. R., Gundel L. A., and Novakov T. (1982) Chemical and catalytic
properties of elemental carbon. In Particulate Carbon: Atmospheric Life Cycle, G. T.
Wolff and R. L. Klimsch, eds., New York, Plenum Press, 15881.
Chang W., Heikes B. G., and Lee M. (2004) Ozone deposition to the sea surface: chemical
enhancement and wind speed dependence. Atmos. Environ. 38, 10539.
Chapman S. (1930) A theory of upper-atmospheric ozone. Mem. Roy. Meteor. Soc. 3, 104
25.
Chapman S. and Cowling T. G. (1970) The Mathematical Theory of Nonuniform Gases.
Cambridge, Cambridge University Press.
Charney J. G. (1949) On a physical basis for numerical prediction of large-scale motions in
the atmosphere. J. Meteor. 6, 37185.
(1951) Dynamical forecasting by numerical process. In Compendium of Meteorology, T. F.
Malone, ed., Boston, American Meteorological Society, 47082.
Charney J. G. and Phillips N. A. (1953) Numerical integration of the quasigeostrophic equa-
tions for barotropic and simple baroclinic ows. J. Meteor. 10, 7199.
Charnock H. (1955) Wind stress on a water surface. Q. J. Roy. Meteor. Soc. 81, 63940.
Chateld R. B., Gardner E. P., and Calvert J. G. (1987) Sources and sinks of acetone in the
troposphere: Behavior of reactive hydrocarbons and a stable product. J. Geophys. Res.
92, 420816.
Chen C. (1991) A nested grid, nonhydrostatic, elastic model using a terrain-following coor-
dinate transformation: The radiative-nesting boundary conditions. Mon. Wea. Rev. 119,
285269.
Chen S., Cope V. W., and Hoffman M. Z. (1973) Behavior of CO
3
radicals generated in the

ash photolysis of carbonatoamines complexes of cobalt(III) in aqueous solution. J. Phys.
Chem. 77, 11116.
Cheng M.-D. and Arakawa A. (1997) Inclusion of rainwater budget and convective down-
drafts in the Arakawa-Schubert cumulus parameterization. J. Atmos. Sci. 54, 1359
78.
Cheng Y., Canuto V. M., and Howard A. M. (2002) An improved model for the turbulent
PBL. J. Atmos. Sci. 59, 155065.
Chock D. P. (1991) A comparison of numerical methods for solving the advection equation
III. Atmos. Environ. 25A, 85371.
756
References
Chock D. P. and Winkler S. L. (1994) A comparison of advection algorithms coupled with
chemistry. Atmos. Environ. 28, 265975.
(2000) A trajectory-grid approach for solving the condensation and evaporation equations
of aerosols. Atmos. Environ. 34, 295773.
Chock D. P., Sun P., and Winkler S. L. (1996) Trajectory-grid: An accurate sign-preserving
advectiondiffusion approach for air quality modeling. Atmos. Environ. 30, 85768.
Christensen H., Sehested K., and Cortzen H. (1982) Reactions of hydroxyl radicals with
hydrogen peroxide at ambient and elevated temperatures. J. Phys. Chem. 86, 158890.
Chylek P. (1977) A note on extinction and scattering efciencies. J. Appl. Meteor. 16,
3212.
Chylek P., Srivastava V., Pinnick R. G., and Wang R. T. (1988) Scattering of electromagnetic
waves by composite spherical particles: experiment and effective mediumapproximations.
Appl. Opt. 27, 2396404.
Chylek P., Videen G., Ngo D., Pinnick R. G., and Klett J. D. (1995) Effect of black carbon
on the optical properties and climate forcing of sulfate aerosols. J. Geophys. Res. 100,
16,32532.
Clapp R. B. and Hornberger G. M. (1978) Empirical equations for some soil hydraulic prop-
erties. Water Resour. Res. 14, 6014.
Cleaver B., Rhodes E., and Ubbelohde A. R. (1963) Studies of phase transformations in nitrates
and nitrites I. Changes in ultra-violet absorption spectra on melting. Proc. Roy. Soc.
London 276, 43753.
Clegg S. L. and Brimblecombe P. (1995) Application of a multicomponent thermodynamic
model to activities and thermal properties of 040 mol kg
1
aqueous sulphuric acid from
- 200 K to 328 K. J. Chem. Eng. Data 40, 4364.
Clegg S. L. and Seinfeld J. H. (2004) Improvement of the ZdanovskiiStokesRobinson
model for mixtures containing solutes of different charge types. J. Phys. Chem. 108,
100817.
Clegg S. L., Brimblecombe P., Liang Z., and Chan C. K. (1997) Thermodynamic properties of
aqueous aerosols to high supersaturation: II A model of the system Na
+
Cl
NO
3
SO
4
2
H
2
O at 298.15 K. Aerosol. Sci. Technol. 27, 34566.
Clegg S. L., Seinfeld J. H., and Edney E. O. (2003) Thermodynamic modeling of aque-
ous aerosols containing electrolytes and dissolved organic compounds. II. An extended
ZdanovskiiStokesRobinson approach. J. Aerosol Sci. 34, 66790.
Clyne M. A. A., Monkhouse P. B., and Townsend L. W. (1976) Reactions of O(
3
P) atoms with
halogens: The rate constants for the elementary reactions O(
3
P) + BrCl, O(
3
P) + Br
2
and
O(
3
P) + Cl
2
. Int. J. Chem. Kinet. 8, 42549.
Coffman D. J. and Hegg D. A. (1995) A preliminary study of the effect of ammonia on particle
nucleation in the marine boundary layer. J. Geophys. Res. 100, 714760.
Cohen M. D., Flagan R. C., and Seinfeld J. H. (1987a) Studies of concentrated electrolyte solu-
tions using the electrodynamic balance. 1. Water activities for single-electrolyte solutions.
J. Phys. Chem. 91, 456374.
(1987b) Studies of concentrated electrolyte solutions using the electrodynamic balance. 2.
Water activities for mixed-electrolyte solutions. J. Phys. Chem. 91, 457582.
Coleman G. N. (1999) Similarity statistics from a direct numerical simulation of the neutrally
stratied planetary boundary layer. J. Atmos. Sci. 56, 8919.
Collela P. and Woodward P. R. (1984) The piecewise parabolic method (PPM) for gas-
dynamical simulations. J. Comp. Phys. 54, 174201.
Comes F. J., Forberich O., and Walter J. (1997) OH eld measurements: A critical input into
model calculations on atmospheric chemistry. J. Atmos. Sci. 54, 188694.
Conklin M. H. and Hoffmann M. R. (1988) Metal ion-S(IV) chemistry III. Thermodynamics
and kinetics of transient iron(III)-sulfur(IV) complexes. Environ. Sci. Technol. 22, 8918.
Cooke W. F. and Wilson J. J. N. (1996) A global black carbon aerosol model. J. Geophys. Res.
101, 19,395409.
757
References
Cooke, W. F., Liousse C., Cachier H., and Feichter J. (1999) Construction of a 1
fossil
fuel emission data set for carbonaceous aerosol and implementation and radiative impact
in the ECHAM4 model. J. Geophys. Res. 104, 22,13762.
Cotton W. R. and Anthes R. A. (1989) Storm and Cloud Dynamics. San Diego, Academic
Press, Inc.
Courant R., Friedrichs K., and Lewy H. (1928)

Uber die partiellen Differenzengleichungen der
mathematischen Physik. Math. Ann. 100, 3274.
Crank J. (1975) The Mathematics of Diffusion, 2nd edn. Oxford, Clarendon Press.
Crank J. and Nicolson P. (1947) A practical method for numerical evaluation of solutions of
partial differential equations of the heat-conduction type. Proc. Camb. Philos. Soc. 43,
5067.
Crutzen P. J. (1971) Ozone production rates in an oxygen-hydrogen-nitrogen oxide atmo-
sphere. J. Geophys. Res. 76, 731127.
Cuenca R. H., Ek M., and Mahrt L. (1996) Impact of soil water property parameterization
on atmospheric boundary layer simulation. J. Geophys. Res. 101, 726977.
Cunningham E. (1910) On the velocity of steady fall of spherical particles through uid
medium. Proc. Roy. Soc. London A83, 35765.
Curtiss C. F. and Hirschfelder J. O. (1952) Integration of stiff equations. Proc. Nat. Acad. Sci.
USA 38, 23543.
Cuzzi J. N., Ackerman T. P., and Helmle L. C. (1982) The delta-four-stream approximation
for radiative transfer. J. Atmos. Sci. 39, 91725.
Dabdub D. and Seinfeld J. H. (1994) Numerical advective schemes used in air quality
models sequential and parallel implementation. Atmos. Environ. 28, 336985.
(1995) Extrapolation techniques used in the solution of stiff ODEs associated with chemical
kinetics of air quality models. Atmos. Environ. 29, 40310.
Danielsen, E. F., Bleck R., and Morris D. A. (1972) Hail growth by stochastic collection in a
cumulus model. J. Atmos. Sci. 29, 13555.
Davies H. C. (1976) A lateral boundary formulation for multi-level prediction models. Q. J.
Roy. Meteor. Soc. 102, 40518.
Davis E. J. (1983) Transport phenomena with single aerosol particles. Aerosol Sci. Technol.
2, 12144.
Dean, J. A. (1992) Langes Handbook of Chemistry. New York, McGraw-Hill, Inc.
Deardorff J. W. (1972) Numerical investigation of neutral and unstable planetary boundary
layers. J. Atmos. Sci. 29, 91115.
(1977) A parameterization of ground surface moisture content for use in atmospheric
prediction models. J. Appl. Meteor. 16, 11825.
(1978) Efcient prediction of ground surface temperature and moisture with inclusion of a
layer of vegetation. J. Geophys. Res. 83, 1889903.
De Arellano J. V., Duynkerke P., and van Weele M. (1994) Tethered-balloon measurements
of actinic ux in a cloud-capped marine boundary layer. J. Geophys. Res. 99, 3699
705.
De Leeuw G., Neele F. P., Hill M., Smith M. H., and Vignati E. (2000) Production of sea spray
aerosol in the surf zone. J. Geophys. Res. 105, 29397409.
Deirmendjian D. (1969) Electromagnetic Scattering on Spherical Polydispersions. New York,
Elsevier.
DeMore W. B., Sanders S. P., Golden D. M., et al. (1997) Chemical Kinetics and Photochem-
ical Data for Use in Stratospheric Modeling. Evaluation number 12, JPL Publ. 97-4, Jet
Propulsion Laboratory, Pasadena, CA.
Dickerson R. R., Kondragunta S., Stenchikov G., Civerolo K. L., Doddridge B. G., and Holben
B. N. (1997) The impact of aerosols on solar UV radiation and photochemical smog.
Science 278, 82730.
Dickinson R. E. (1984) Modeling evapotranspiration for three-dimensional global climate
models. In Climate Processes and Climate Sensitivity, Geophys. Monogr. Ser., Vol. 29,
758
References
J. E. Hanson and T. Takahashi, eds., Washington, DC, American Geophysical Union,
5872.
Ding P. and Randall D. A. (1998) A cumulus parameterization with multiple cloud-base levels.
J. Geophys. Res. 103, 11,34153.
Donea J. (1984) A TaylorGalerkin method for convective transport problems. Int. J. Numer.
Methods Engng. 20, 10119.
Dorsch R. G. and Hacker P. (1951) Experimental Values of Surface Tension of Supercooled
Water. National Advisory Committee for Aeronautics (NACA), Tech. Note 2510.
Duce R. A. (1969) On the source of gaseous chlorine in the marine atmosphere. J. Geophys.
Res. 70, 17759.
Dudhia J. (1993) A nonhydrostatic version of the Penn State-NCAR mesoscale model:
Validation tests and simulation of an Atlantic cyclone and cold front. Mon. Wea. Rev.
121, 1493513.
Durran D. R. (1999) Numerical Methods for Wave Equations in Geophysical Fluid Dynamics.
New York, Springer-Verlag.
Dyer A. J. (1974) A review of ux-prole relationships. Boundary-Layer Meteor. 7, 363
72.
Dyer A. J. and Bradley E. F. (1982) An alternative analysis of ux-gradient relationships at the
1976 ITCE. Boundary-Layer Meteor. 22, 319.
Easter R. C. (1993) Two modied versions of Botts positive-denite numerical advection
scheme. Mon. Wea. Rev. 121, 297304.
Eddington S. A. (1916) On the radiative equilibrium of the stars. Mon. Not. Roy. Astron. Soc.
77, 1635.
Edlen B. (1966) The refractive index of air. Meteorology 2, 7180.
Eliasen E., Machenhauer B., and Rasmussen E. (1970) On a Numerical Method for Integra-
tion of the Hydrodynamical Equations with a Spectral Representation of the Horizon-
tal Fields. Report No. 2, Institut for Teoretisk Meteorologi, University of Copenhagen,
35pp.
Elliott D. L., Holladay C. G., Barchet W. R., Foote H. P., and Sandusky W. F. (1986) Wind
Energy Resource Atlas of the United States. DOE/CH10093-4 Natl. Renew. Energy Lab.,
Golden, CO.
Elliott S., Turco R. P., and Jacobson M. Z. (1993) Tests on combined projection/forward dif-
ferencing integration for stiff photochemical family systems at long time step. Computers
Chem. 17, 91102.
Emanuel K. A. (1991) A scheme for representing cumulus convection in large-scale models. J.
Atmos. Sci. 38, 154157.
Enger L. (1986) A higher order closure model applied to dispersion in a convective PGL.
Eriksson E. (1960) The yearly circulation of chloride and sulfur in nature; meteorological,
geochemical and pedological implications. Part II. Tellus 12, 63109.
Erisman J. W., van Pul W. A. J., and Wyers P. (1994) Parameterization of surface resistance for
the quantication of atmospheric deposition of acidifying pollutants and ozone. Atmos.
Environ. 28, 2595607.
Facchini M. C., Mircea M., Fuzzi S., and Charlson R. J. (1999) Cloud albedo enhancement
by surface-active organic solutes in growing droplets. Nature 401, 2579.
Fang, M., Zheng M., Wang F., To K. L., Jaafar A. B., and Tong S. L. (1999) The solvent-
extractable organic compounds in the Indonesia biomass burning aerosols characteri-
zation studies. Atmos. Environ. 33, 78395.
Farhataziz and Ross A. B. (1977) Selected Specic Rates of Transients From Water in Aqueous
Solutions, III. Hydroxyl Radical and Perhydroxyl Radical and Their Radical Ions, Rep.
NSRDBS-NBS 59, U.S. Department of Commerce, Washington, DC.
Farman J. C., Gardiner B. G., and Shanklin J. D. (1985). Large losses of total ozone in Antarc-
tica reveal seasonal ClO
x
/NO
x
interaction. Nature 315, 20710.
759
References
Fassi-Fihri A., Suhre K., and Rosset R. (1997) Internal and external mixing in atmospheric
aerosols by coagulation: Impact on the optical and hygroscopic properties of the sulphate
soot system. Atmos. Environ. 31, 1392402.
Ferek, R. J., Reid J. S., Hobbs P. V., Blake D. R., and Liousse C. (1998) Emission factors of
hydrocarbons, halocarbons, trace gases, and particles from biomass burning in Brazil.
J. Geophys. Res. 103, 32,10718.
Fernandez-Diaz J. M., Gonzalez-Pola Muniz C., Rodriguez Brana M. A., Arganza Garcia B.,
and Garcia Nieto P. J. (2000) A modied semi-implicit method to obtain the evolution of
an aerosol by coagulation. Atmos. Environ. 34, 430114.
Filippov V. K., Charykova M. V., and Tromov Y. M. (1985) Thermodynamics of the system
NH
4
H
2
PO
4
-(NH
4
)
2
SO
4
-H
2
O at 25

C. J. Appl. Chem. USSR 58, 180711.
Finlayson-Pitts B. and Pitts Jr. J. N. (2000) Chemistry of the Upper and Lower Atmosphere,
San Diego, Academic Press.
Fleming E. L., Chandra S., Schoeberl M. R., and Barnett J. J. (1988) Monthly Mean Global
Climatology of Temperature, Wind, Geopotential Height, and Pressure for 1120 km.
Tech. Memo. 100697, NASA, 85pp.
Fletcher N. H. (1958) Size effect in heterogeneous nucleation. J. Chem. Phys. 29, 57276.
Flossmann A. I., Hall W. D., and Pruppacher H. R. (1985) A theoretical study of the wet
removal of atmospheric pollutants. Part I: The redistribution of aerosol particles captured
through nucleation and impaction scavenging by growing cloud drops. J. Atmos. Sci. 42,
582606.
Flubacher P., Leadbetter A. J., and Morrison J. A. (1960) Heat capacity of ice at low temper-
atures. J. Chem. Phys. 33, 17515.
Foster V. G. (1992) Determination of the refractive index dispersion of liquid nitrobenzene in
the visible and ultraviolet. J. Phys. D 25, 5259.
Fowler L. D., Randall D. A., and Rutledge S. (1996) Liquid and ice cloud microphysics in the
CSU general circulation model. Part I: Model description and simulated microphysical
processes. J. Climate 9, 489529.
Frank W. M. and Cohen C. (1987) Simulation of tropical convective systems. Part I: Acumulus
parameterization. J. Atmos. Sci. 44, 378799.
Freedman F. R. and Jacobson M. Z. (2002) Transport-dissipation analytical solutions to the
E- turbulence model and their role in predictions of the neutral ABL. Boundary-Layer
Meteor. 102, 11738.
(2003) Modication of the standard -equation for the stable ABL through enforced consis-
tency with MoninObukhov similarity theory. Boundary-Layer Meteor. 106, 383410.
Fridlind A. M. and Jacobson M. Z. (2000) A study of gasaerosol equilibrium and aerosol pH
in the remote marine boundary layer during the First Aerosol Characterization Experiment
(ACE 1). J. Geophys. Res. 105, 1732540.
(2003) Point and column aerosol radiative closure during ACE 1: Effects of particle shape
and size. J. Geophys. Res. 108 (D3) doi:10.1029/2001JD001553.
Friedlander S. K. (1977) Smoke, Dust, and Haze. Fundamentals of Aerosol Behavior. New
York, John Wiley & Sons, Inc.
(1983) Dynamics of aerosol formation by chemical reaction. Ann. N. Y. Acad. Sci. 404,
35464.
Friedli H., L otscher H., Oeschger H., Siegenthaler U., and Stauffer B. (1996) Ice core record
of 13C/12C ratio of atmospheric CO
2
in the past two centuries. Nature 324, 2378.
Fritsch J. M. and Chappel C. F. (1980) Numerical prediction of convectively driven mesoscale
pressure systems. Part I: Convective parameterization. J. Atmos. Sci. 37, 172233.
Fu Q. and Liou K. N. (1992) On the correlated k-distribution method for radiative transfer
in nonhomogeneous atmospheres. J. Atmos. Sci. 49, 213956.
(1993) Parameterization of the radiative properties of cirrus clouds. J. Atmos. Sci. 50, 2008
25.
760
References
Fu Q., Liou K. N., Cribb M. C., Charlock T. P., and Grossman A. (1997) Multiple scattering
parameterization in thermal infrared radiative transfer. J. Atmos. Sci 54, 2799812.
Fuchs N. A. (1964) The Mechanics of Aerosols (translated by R. E. Daisley and M. Fuchs).
New York, Pergamon Press.
Fuchs N. A. and Sutugin A. G. (1971). Highly dispersed aerosols. In Topics in Current
Aerosol Research, Vol. 2, G. M. Hidy and J. R. Brock, eds., New York, Pergamon Press,
160.
Fuller K. A. (1995) Scattering and absorption cross sections of compounded spheres. III.
Spheres containing arbitrarily located spherical inhomogeneities. J. Opt. Soc. Am. A 12,
893904.
Fuller K. A., Malm W. C., and Kreidenweis S. M. (1999) Effects of mixing on extinction by
carbonaceous particles. J. Geophys. Res. 104, 15,94154.
Galbally I. E. and Roy C. R. (1980) Destruction of ozone at the earths surface. Q. J. Roy.
Meteor. Soc. 106, 599620.
Ganzeveld L. and Lelieveld J. (1995) Dry deposition parameterization in a chemistry general
circulation model and its inuence on the distribution of reactive trace gases. J. Geophys.
Res. 100, 20,9991,012.
Garcia R. R., Stordal F., Solomon S., and Kiehl J. T. (1992) A new numerical model of the
middle atmosphere 1. Dynamics and transport of tropospheric source gases. J. Geophys.
Res. 97, 12,96791.
Garratt J. R. (1992) The Atmospheric Boundary Layer. Cambridge, Cambridge University
Press.
Garratt J. R. and Hicks B. B. (1973) Momentum, heat and water vapour transfer to and from
natural and articial surfaces. Q. J. Roy. Meteor. Soc. 99, 6807.
Gaydos, T. M., Koo B., Pandis S. N., and Chock D. P. (2003) Atmos. Environ. 37, 330316.
Gazdag J. (1973) Numerical convective schemes based on accurate computation of space
derivatives. J. Comp. Phys. 13, 10013.
Gear C. W. (1971) Numerical Initial Value Problems in Ordinary Differential Equations.
Englewood Cliffs, NJ, Prentice-Hall.
Gelbard F. (1990) Modeling multicomponent aerosol particle growth by vapor condensation.
Gelbard F. and Seinfeld J. H. (1980) Simulation of multicomponent aerosol dynamics. J. Col-
loid Interface Sci. 78, 485501.
Gelbard F., Fitzgerald J. W., and Hoppel W. A. (1998) A one-dimensional sectional model to
simulate multicomponent aerosol dynamics in the marine boundary layer. 3. Numerical
methods and comparisons with exact solutions. J. Geophys. Res. 103, 16,119132.
Gerber H., Takano Y., Garrett T. J., and Hobbs P. V. (2000) Nephelometer measurements of
the asymmetry parameter, volume extinction coefcient, and backscatter ratio in Arctic
clouds. J. Atmos. Sci. 57, 302133.
Gery M. W., Whitten G. Z., and Killus J. P. (1988) Development and Testing of the CBM-
IV for Urban and Regional Modeling. Report EPA-600/3-88-012. U.S. Environmental
Protection Agency, Research Triangle Park, NC.
Gery M. W., Whitten G. Z., Killus J. P., and Dodge M. C. (1989) A photochemical kinetics
mechanismfor urban and regional scale computer modeling. J. Geophys. Res. 94, 12,925
56.
Ghio A. J. and Samet J. M. (1999) Metals and air pollution particles. In Air Pollution and
Health, S. T. Holgate, J. M. Samet, H. S. Koren, and R. L. Maynard, eds., San Diego,
Academic Press, 63551.
Giauque W. F. and Stout J. W. (1936) The entropy of water and the third law of thermody-
namics. The heat capacity of ice from 15 to 273 K. J. Am. Chem. Soc. 58, 114450.
Gillette D. A. (1974) On the production of soil wind erosion aerosols having the potential for
long range transport. Atmos. Res. 8, 73544.
761
References
Gillette D. A., Patterson Jr. E. M., Prospero J. M., and Jackson M. L. (1993) Soil aerosols.
In Aerosol Effects on Climate, S. G. Jennings, ed., Tucson, University of Arizona Press,
73109.
Giorgi F., Marinucci M. R., Bates G. T., and De Canio G. (1993) Development of a second-
generation climate model (RegCM2) Part II: Convective processes and assimilation of
lateral boundary conditions. Mon. Wea. Rev. 121, 281432.
Gittens G. J. (1969) Variation of surface tension of water with temperature. J. Colloid Interface
Sci. 30, 40612.
Goldberg R. N. (1981) Evaluated activity and osmotic coefcients for aqueous solutions:
Thirty-six uni-bivalent electrolytes. J. Phys. Chem. Ref. Data 10, 671764.
Golding B. W. (1992) An efcient nonhydrostatic forecast model. Meteor. Atmos. Phys. 50,
89103.
Gong W. and Cho H.-R. (1993) A numerical scheme for the integration of the gas-phase
chemical rate equations in three-dimensional atmospheric models. Atmos. Environ. 27A,
214760.
Goodin W. R., McRae G. J., and Seinfeld J. H. (1979) Acomparison of interpolations methods
for sparse data: Application to wind and concentration elds. J. Appl. Meteor. 18, 76171.
Goody, R. M., West R., Chen L., and Crisp D. (1989) The correlated-k method for radiation
calculations in nonhomogeneous atmospheres. J. Quant. Spectrosc. Radiat. Transfer 42,
53950.
Graedel T. E. and Goldberg K. I. (1983) Kinetic studies of raindrop chemistry, 1. Inorganic
and organic processes. J. Geophys. Res. 88, 10,86582.
Graedel T. E. and Weschler C. J. (1981) Chemistry within aqueous atmospheric aerosols and
raindrops. Rev. Geophys. 19, 50539.
Greeley R. and Iversen J. D. (1985) Wind as a Geological Process on Earth, Mars, Venus, and
Titan. New York, Cambridge University Press, 333pp.
Greenberg R. R., Zoller W. H., and Gordon G. E. (1978) Composition and size distributions
of articles released in refuse incineration. Environ. Sci. Technol. 12, 56673.
Grell G. A. (1993) Prognostic evaluation of assumptions used by cumulus parameterizations.
Mon. Wea. Rev. 121, 76487.
Grifn R. J., Dabdub D., and Seinfeld J. H. (2002) Secondary organic aerosol 1. Atmospheric
chemical mechanism for production of molecular constituents. J. Geophys. Res. 107
(D17), 4332, doi:10.1029/2001JD000541.
Groblicki P. J., Wolff G. T., and Countess R. J. (1981) Visibility-reducing species in the Denver
brown cloud I. Relationships between extinction and chemical composition. Atmos.
Environ. 15, 247384.
Guelle W., Schulz M., Balkanski Y., and Dentener F. (2001) Inuence of source formulation
on modeling the atmospheric global distribution of sea salt aerosol. J. Geophys. Res. 106,
2750924.
Hack J. J. (1992) Climate system simulation: Basic numerical and computational concepts. In
Climate System Modeling, K. E. Trenberth, ed., Cambridge, Cambridge University Press,
283318.
(1994) Parameterization of moist convection in the National Center for Atmospheric
Research community climate model (CCM2). J. Geophys. Res. 99, 555168.
Hagesawa K. and Neta P. (1978) Rate constants and mechanisms of reaction for Cl
2
radicals.
J. Phys. Chem. 82, 8547.
Hairer E. and Wanner G. (1991) Solving Ordinary Differential Equations II. Stiff and
Differential-Algebraic Problems. Berlin, Springer-Verlag.
Hale G. M. and Querry M. R. (1973) Optical constants of water in the 200-nm to 200-m
wavelength region. Appl. Opt. 12, 55563.
Hamer W. J. and Wu Y.-C. (1972) Osmotic coefcients and mean activity coefcients of uni-
univalent electrolytes in water at 25

C. J. Phys. Chem. Ref. Data 1, 104799.
762
References
Hamill P., Turco R. P., Kiang C. S., Toon O. B., and Whitten R. C. (1982) An analysis of
various nucleation mechanisms for sulfate particles in the stratosphere. J. Aerosol Sci. 13,
56185.
Hansen J. E. (1969) Radiative transfer by doubling very thin layers. Astrophys. J. 155, 565
73.
Harned H. S. and Owen B. B. (1958) The Physical Chemistry of Electrolyte Solutions. New
York, Reinhold, Chapter 8.
Harrington J. Y., Meyers M. P., Walko R. L., and Cotton W. R. (1995) Parameterization of ice
crystal conversion process due to vapor deposition for mesoscale models using double-
moment basis functions. Part I: Basic formulation and parcel model results. J. Atmos. Sci.
52, 434466.
Harris S. J. and Maricq, M. M. (2001) Signature size distributions for diesel and gasoline
engine exhaust particulate matter. J. Aerosol Sci. 32, 74964.
Hartmann D. L. (1994) Global Physical Climatology. San Diego, Academic Press, Inc.
Harvey R. B., Stedman D. H., and Chameides W. (1977) Determination of the absolute rate
of solar photolysis of NO
2
. J. Air Pol. Control Assn. 27, 6636.
Harvie C. E., Moller N., and Weare J. H. (1984) The prediction of mineral solubilities in
natural waters: The Na-K-Mg-Ca-H-Cl-SO
4
-OH-HCO
3
-CO
3
-CO
2
-H
2
O system to high
ionic strengths at 25

C. Geochim. Cosmochim. Acta 48, 72351.
Henry W. M. and Knapp K. T. (1980) Compound forms of fossil fuel y ash emissions. Environ.
Sci. Technol. 14, 4506.
Henyey L. C. and Greenstein J. L. (1941) Diffuse radiation in the galaxy. Astrophys. J. 93,
7083.
Hering S. V. and Friedlander S. K. (1982) Origins of aerosol sulfur size distributions in the
Los Angeles Basin. Atmos. Environ. 16, 264756.
Hering S. V., Friedlander S. K., Collins J. C., and Richards L. W. (1979) Design and evaluation
of a new low pressure impactor 2. Environ. Sci. Technol. 13, 1848.
Hertel O., Berkowicz R., and Christensen J. (1993) Test of two numerical schemes for use in
atmospheric transport-chemistry models. Atmos. Environ. 27A, 2591611.
Hesstvedt E., Hov O., and Isaksen I. S. A. (1978) Quasi-steady-state approximations in air
pollution modeling: Comparison of two numerical schemes for oxidant prediction. Int.
J. Chem. Kin. 10, 97194.
Himmelblau P. M. (1964) Diffusion of dissolved gases in liquids. Chem. Rev. 64, 52750.
Hindmarsh A. C. (1983) ODEPACK, a systematized collection of ODE solvers. In Scientic
Computing, R. S. Stepleman et al., eds., Amsterdam, North-Holland, 5574.
Hinze J. O. (1975) Turbulence: An Introduction to its Mechanism and Theory, 2nd edn. New
York, McGraw-Hill, 790pp.
Hitchcock D. R., Spiller L. L., and Wilson W. E. (1980) Sulfuric acid aerosols and HCl release
in coastal atmospheres: Evidence of rapid formation of sulfuric acid particulates. Atmos.
Environ. 14, 16582.
Hoffmann M. R. and Calvert J. G. (1985) Chemical Transformation Modules for Eulerian
Acid Deposition Models, Vol. 2. The Aqueous-phase Chemistry. EPA/600/3-85/017. U.S.
Environmental Protection Agency, Research Triangle Park, NC.
Hogstrom U. (1988) Non-dimensional wind and temperature proles in the atmospheric sur-
face layer: A reevaluation. Boundary-Layer Meteor. 42, 5578.
Holmes H. F. and Mesmer R. E. (1986) Thermodynamics of aqueous solutions of the alkali
metal sulfates. J. Solution Chem. 15, 495518.
Holton J. R. (1992) An Introduction to Dynamical Meteorology. San Diego, Academic Press,
Inc.
Hounslow M. J., Ryall R. L., and Marshall V. R. (1988) A discretized population balance for
nucleation, growth, and aggregation. AIChE J. 34, 182132.
Houze R. A., Jr. (1993) Cloud Dynamics. San Diego, Academic Press, Inc.
763
References
Huang H.C. and Chang J. S. (2001) On the performance of numerical solvers for a chemistry
submodel in three-dimensional air quality models. 1. Box model simulations. J. Geophys.
Res. 106, 20,17588.
Hughes T. J. R. and Brooks A. N. (1979) Amultidimensional upwind scheme with no crosswind
diffusion. In Finite Element Methods for Convection Dominated Flows, T. J. R. Hughes,
ed., AMD Vol. 34, ASME, New York, 1935.
Huie R. E. and Neta P. (1987) Rate constants for some oxidations of S(IV) by radicals in
aqueous solutions. Atmos. Environ. 21, 17437.
Hulburt H. M. and Katz S. (1964) Some problems in particle technology: Astatistical mechan-
ical formulation. Chem. Eng. Sci. 19, 55574.
Hynes A. J., Wine P. H., and Semmes D. H. (1986) Kinetic mechanism of OH reactions with
organic suldes. J. Phys. Chem. 90, 414856.
Irvine W. M. (1968) Multiple scattering by large particles. II. Optically thick layers. Astrophys.
J. 152, 82334.
(1975) Multiple scattering in planetary atmospheres. Icarus 25, 175204.
Ishizaka Y. and Adhikari M. (2003) Composition of cloud condensation nuclei. J. Geophys.
Res. 108 (D4), 4138, doi:10.1029/2002JD002085.
Jackman C. H., Fleming E. L., Chandra S., Considine D. B., and Roseneld J. E. (1996)
Past, present, and future modeled ozone trends with comparisons to observed trends.
J. Geophys. Res. 101, 28,75367.
Jacob D. J. (1986) Chemistry of OH in remote clouds and its role in the production of formic
acid and peroxymonosulfonate. J. Geophys. Res. 91, 980726.
Jacob D. J., Gottlieb E. W., and Prather M. J. (1989a) Chemistry of a polluted cloudy boundary
layer. J. Geophys. Res. 94, 12,97513,002.
Jacob D. J., Sillman S., Logan J. A., and Wofsy S. C. (1989b) Least independent variables
method for simulation of tropospheric ozone. J. Geophys. Res. 94, 8497509.
Jacobson M. Z. (1994) Developing, coupling, and applying a gas, aerosol, transport, and radi-
ation model to study urban and regional air pollution. Ph.D. Thesis, Dept. of Atmospheric
Sciences, University of California, Los Angeles.
(1995) Computation of global photochemistry with SMVGEAR II. Atmos. Environ. 29A,
25416.
(1997a) Development and application of a new air pollution modeling system. Part II:
Aerosol module structure and design. Atmos. Environ. 31A, 13144.
(1997b) Development and application of a new air pollution modeling system. Part III:
Aerosol-phase simulations. Atmos. Environ. 31A, 587608.
(1997c) Numerical techniques to solve condensational and dissolutional growth equa-
tions when growth is coupled to reversible aqueous reactions. Aerosol Sci. Technol. 27,
4918.
(1998a) Vector and scalar improvement of SMVGEAR II through absolute error tolerance
control. Atmos. Environ. 32, 7916.
(1998b) Studying the effects of aerosols on vertical photolysis rate coefcient and tempera-
ture proles over an urban airshed. J. Geophys. Res. 103, 10,593604.
(1999a) Effects of soil moisture on temperatures, winds, and pollutant concentrations in
Los Angeles. J. Appl. Meteorol. 38, 60716.
(1999b) Studying the effects of calcium and magnesium on size-distributed nitrate and
ammonium with EQUISOLV II. Atmos. Environ. 33, 363449.
(1999c) Isolating nitrated and aromatic aerosols and nitrated aromatic gases as sources of
ultraviolet light absorption. J. Geophys. Res. 104, 352742.
(2000) A physically-based treatment of elemental carbon optics: Implications for global
direct forcing of aerosols. Geophys. Res. Lett. 27, 21720.
(2001a) GATOR-GCMM: A global- through urban-scale air pollution and weather forecast
model 1. Model design and treatment of subgrid soil, vegetation, roads, rooftops, water,
sea ice, and snow. J. Geophys. Res. 106, 5385401.
764
References
(2001b) Strong radiative heating due to the mixing state of black carbon in atmospheric
aerosols. Nature 409, 6957.
(2002) Analysis of aerosol interactions with numerical techniques for solving coagulation,
nucleation, condensation, dissolution, and reversible chemistry among multiple size dis-
tributions. J. Geophys. Res. 107 (D19), 4366, doi:10.1029/2001JD002044.
(2003) Development of mixed-phase clouds from multiple aerosol size distributions
and the effect of the clouds on aerosol removal. J. Geophys. Res. 108 (D8), 4245,
doi:10.1029/2002JD002691.
(2005a) A rened method of parameterizing absorption coefcients among multiple gases
simultaneously from line-by-line data. J. Atmos. Sci. 62, 50617.
(2005b) Studying ocean acidication with conservative, stable numerical schemes for
nonequilibrium airocean exchange and ocean equilibrium chemistry. J. Geophys. Res.,
in review.
(2005c) A solution to the problem of nonequilibrium acid/base gas-particle transfer at long
time step. Aerosol Sci. Technol. 39, 92103.
Jacobson M. Z. and Seinfeld J. H. (2004) Evolution of nanoparticle size and mixing state near
the point of emission. Atmos. Environ. 38, 183950.
Jacobson M. Z. and Turco R. P. (1994) SMVGEAR: A sparse-matrix, vectorized Gear code
for atmospheric models. Atmos. Environ. 28A, 27384.
(1995) Simulating condensational growth, evaporation, and coagulation of aerosols using
a combined moving and stationary size grid. Aerosol Sci. Technol. 22, 7392.
Jacobson M. Z., Turco R. P., Jensen E. J., and Toon O. B. (1994) Modeling coagulation among
particles of different composition and size. Atmos. Environ. 28A, 132738.
Jacobson M. Z., Tabazadeh A., and Turco R. P. (1996b) Simulating equilibriumwithin aerosols
and non-equilibrium between gases and aerosols. J. Geophys. Res. 101, 907991.
Jaecker-Voirol A. and Mirabel P. (1989) Heteromolecular nucleation in the sulfuric acid-water
system. Atmos. Environ. 23, 203357.
Jaenicke R. (1988) Aerosol physics and chemistry. In Numerical Data and Functional Rela-
tionships in Science and Technology. New Series Vol. 4, Meteorology Subvol. b, Physical
and Chemical Properties of Air, G. Fischer, ed., Berlin, Springer-Verlag.
Jarvis P. G., James G. B., and Landsberg J. J. (1976) Coniferous forest. In Vegetation and the
Atmosphere, Vol. 2. J. L. Monteight, ed., New York, Academic Press, 171240.
Jayne J. T., Davidovits P., Worsnop D. R., Zahniser M. S., and Kolb C. E. (1990) Uptake
of SO
2
by aqueous surfaces as a function of pH: The effect of chemical reaction at the
interface. J. Phys. Chem. 94, 60418.
Jayson G. G., Parsons B. J., and Swallow A. J. (1973) Some simple, highly reactive, inorganic
chlorine derivatives in aqueous solution. Trans. Faraday Soc. 69, 1597607.
Jeans J. (1954) The Dynamical Theory of Gases. New York, Dover.
Jenkin M. E., Saunders S. M., Wagner V., and Pilling J. (2003) Protocol for the development
of the Master Chemical Mechanism, MCM v3 (Part B): tropospheric degradation of
aromatic volatile organic compounds. Atmos. Chem. Phys. 3, 18193.
John W., Wall S. M., Ondo J. L., and Winklmayr W. (1989) Acidic Aerosol Size Distributions
During SCAQS. Final Report for the California Air Resources Board under Contract No.
A6-112-32.
Joseph J. H., Wiscombe W. J., and Weinman J. A. (1976) The delta-Eddington approximation
for radiative ux transfer. J. Atmos. Sci. 33, 24529.
Joslin R. D., Streett C. L., and Chang C.-L. (1993) Spatial direct numerical simulation of
boundary-layer transition mechanismsvalidation of PSE theory. Theor. Comput. Fluid
Dyn. 4, 27188.
Junge C. E. (1961) Vertical proles of condensation nuclei in the stratosphere. J. Meteor. 18,
5019.
Kaimal J. C. and Finnigan J. J. (1994) Atmospheric Boundary Layer Flows: Their Structure
and Measurement. New York, Oxford University Press.
765
References
Kain J. S. (2004) The KainFritsch convective parameterization: An update. J. Appl. Meteor.
43, 17081.
Kain J. S. and Fritsch J. M. (1990) A one-dimensional entraining/detraining plume
model and its application in convective parameterization. J. Atmos. Sci. 47, 2784
802.
Kao C.-Y. J. and Ogura Y. (1987) Response of cumulus clouds to large-scale forcing using the
ArakawaSchubert cumulus parameterization. J. Atmos. Sci. 44, 2437548.
Kaps P. and Rentrop P. (1979) Generalized RungeKutta methods of order four with stepsize
control for stiff ordinary differential equations. Numer. Math. 33, 5588.
Kasahara A. (1974) Various vertical coordinate systems used for numerical weather prediction.
Mon. Wea. Rev. 102, 50922.
Kasten F. (1968) Falling speed of aerosol particles. J. Appl. Meteor. 7, 9447.
Katrinak K. A., Rez P., Perkes P. R., and Buseck P. R. (1993) Fractal geometry of carbonaceous
aggregates from an urban aerosol. Environ. Sci. Technol. 27, 53947.
Kawata Y. and Irvine W. M. (1970) The Eddington approximation for planetary atmospheres.
Astrophys. J. 160, 78790.
Keeling C. D. and Whorf T. P. (2003) Atmospheric CO
2
concentrations (ppmv)
derived from in situ air samples collected at Mauna Loa Observatory, Hawaii.
cdiac.esd.ornl.gov./ftp/maunaloa-co2/maunaloa.co2.
Kerker M. (1969) The Scattering of Light and Other Electromagnetic Radiation. New York,
Academic Press.
Ketean G. (2005) Development, testing, and application of a 3-D nonhydrostatic potential-
energy-conserving, compressible atmospheric model. Ph.D. Dissertation, Stanford Uni-
versity.
Ketean G. and Jacobson M. Z. (2005a) Development and application of a 2-D potential-
enstrophy-, energy-, and mass-conserving mixed-layer ocean model with arbitrary bound-
aries, Mon. Wea. Rev., in submission.
(2005b) Development and application of an orthogonal-curvilinear-grid nonhydrostatic
potential-enstrophy-, energy-, and mass-conserving atmospheric model with arbitrary
boundaries, Mon. Wea. Rev., in submission.
Kettle A. J. and Andreae M. O. (2000) Flux of dimethylsulde from the oceans: A comparison
of updated data sets and ux models. J. Geophys. Res. 105, 26,793808.
KimY. P. and Seinfeld J. H. (1995) Atmospheric gas-aerosol equilibrium: III. Thermodynamics
of crustal elements Ca
2+
, K
+
, and Mg
2+
. Aerosol Sci. Technol. 22, 93110.
Kim J., Moin P., and Moser R. (1987) Turbulence statistics in fully developed channel ow at
low Reynolds number. J. Fluid Mech. 177, 13366.
Kim Y. P., Seinfeld J. H., and Saxena P. (1993a) Atmospheric gasaerosol equilibrium I. Ther-
modynamic model. Aerosol Sci. Technol. 19, 15781.
(1993b) Atmospheric gasaerosol equilibrium II. Analysis of common approximations and
activity coefcient calculation methods. Aerosol Sci. Technol. 19, 18298.
Kittelson D. B. (1998) Engine and nanoparticles: a review. J. Aerosol Sci. 6, 44351.
Klemp J. B. and Wilhelmson R. B. (1978) The simulation of three-dimensional convective
storm dynamics. J. Atmos. Sci. 3, 107096.
Klingen H.-J. and Roth P. (1989) Size analysis and fractal dimension of diesel particles
based on REM measurements with an automatic imaging system. J. Aerosol Sci. 20,
8614.
Knudsen M. and Weber S. (1911) Luftwiderstand gegen die langsame Bewegung kleiner
Kugeln. Ann. Phys. 36, 98194.
K ohler H. (1936) The nucleus in the growth of hygroscopic droplets. Trans. Faraday Soc. 32,
115261.
Kondo J., Saigusa N., and Sato T. (1990) A parameterization of evaporation from bare soil
surfaces. J. Appl. Meteor. 29, 3859.
Kondratyev K. Ya. (1969) Radiation in the Atmosphere. San Diego, Academic Press, 912pp.
766
References
Korhonen H., Napari I., Timmreck C., et al. (2003) Heterogeneous nucleation as a poten-
tial sulphate-coating mechanism of atmospheric mineral dust particles and implica-
tions of coated dust on new particle formation. J. Geophys. Res. 108 (D17), 4546,
doi:10.1029/2003JD003553.
Koschmieder H. (1924) Theorie der horizontalen Sichtweite. Beitr. Phys. Freien Atm. 12,
3353, 17181.
Kozac-Channing L. F. and Heltz G. R. (1983) Solubility of ozone in aqueous solutions of 00.6
M ionic strength at 530

C. Environ. Sci. Technol. 17, 1459.
Kreidenweis S. M., Zhang Y., and Taylor G. R. (1997) The effects of clouds on aerosol and
chemical species production and distribution 2. Chemistry model description and sensi-
tivity analysis. J. Geophys. Res. 102, 23,86782.
Kreidenweis S. M., Walcek C., Kim C.-H., et al. (2003) Modication of aerosol mass and
size distribution due to aqueous-phase SO
2
oxidation in clouds: comparison of several
models. J. Geophys. Res. 108 (D7) doi:10.1029/2002JD002697.
Kreitzberg C. W. and Perkey D. (1976) Release of potential instability. Part I: A sequential
plume model within a hydrostatic primitive equation model. J. Atmos. Sci. 33, 45675.
Krekov G. M. (1993) Models of atmospheric aerosols. In Aerosol Effects on Climate. S. G.
Jennings, ed., Tucson, University of Arizona Press, 972.
Krishnamurti T. N. and Moxim W. J. (1971) On parameterization of convective and non-
convective latent heat release. J. Appl. Meteor. 10, 313.
Krishnamurti T. N., Bedi H. S., and Hardiker V. M. (1998) An Introduction to Global Spectral
Modeling. New York, Oxford University Press.
Krishnamurti T. N., Pan H.-L., Pasch R. J., and Molinari J. (1980) Cumulus parameterization
and rainfall rates I. Mon. Wea. Rev. 108, 46572.
Kritz M. A. and Rancher J. (1980) Circulation of Na, Cl, and Br in the tropical marine
atmosphere. J. Geophys. Res. 85, 16339.
Kulmala M., Laaksonen A., and Pirjola L. (1998) Parameterizations for sulfuric acid/water
nucleation rates. J. Geophys. Res. 103, 83017.
Kuo H. L. (1965) On formation and intensication of tropical cyclones through latent heat
release by cumulus convection. J. Atmos. Sci. 22, 4063.
(1974) Further studies of the parameterization of the inuence of cumulus convection on
large-scale ow. J. Atmos. Sci. 31, 123240.
Kurihara Y. (1973) A scheme of moist convective adjustment. Mon. Wea. Rev. 101, 54753.
Kurihara Y. and Bender M. A. (1983) A numerical scheme to treat the open lateral boundary
of limited area model. Mon. Wea. Rev. 111, 44554.
Kusik C. L. and Meissner H. P. (1978) Electrolyte activity coefcients in inorganic processing.
AIChE J. Symp. Ser. 173, 1420.
Lacis A. A. and Hansen J. E. (1974) A parameterization for the absorption of solar radiation
in the Earths atmosphere. J. Atmos. Sci. 31, 11833.
Lacis A., Wang W. C., and Hansen J. (1979) Correlated k-distribution method for radiative
transfer in climate models: Application to effect of cirrus clouds on climate. NASA Conf.
Publ. 2076, 309314.
Lacis A. A. and Oinas V. (1991) Adescription of the correlated k-distribution method for mod-
eling nongray gaseous absorption, thermal emission, and multiple scattering in vertically
inhomogeneous atmospheres. J. Geophys. Res. 96, 902763.
Lamb H. (1910) On atmospheric oscillations. Proc. Roy. Soc. London 84, 55172.
Langford A. O., Proftt M. H., VanZandt T. E., and Lamarque J.-F. (1996) Modulation of
tropospheric ozone by a propagating gravity wave. J. Geophys. Res. 101, 26, 60513.
Larson S., Cass G., Hussey K., and Luce F. (1984) Visibility Model Verication by Image Pro-
cessing Techniques. Final report to the California Air Resources Board under Agreement
A2-077-32.
Lary D. J. (1997) Catalytic destruction of stratospheric ozone. J. Geophys. Res. 102, 21,515
26.
767
References
Lazrus A. L., Cadle R. D., Gandrud B. W., Greenberg J. P., Huebert B. J., and Rose W. I.
(1979) Sulfur and halogen chemistry of the stratosphere and of volcanic eruption plumes.
Ledbury W. and Blair E. W. (1925) The partial formaldehyde vapour pressure of aqueous
solutions of formaldehyde, II. J. Chem. Soc. 127, 283239.
Lee H. D. P., translator (1951) Meteorologica by Aristotle, T. E. Page, ed., Cambridge, MA,
Harvard University Press.
Lee K. O., Cole R., Sekar R., et al. (2001) Detailed characterization of morphology and
dimensions of diesel particulates via thermophoretic sampling, SAF 2001-01-3572.
Lee K. W. (1985) Conservation of particle size distribution parameters during Brownian coag-
ulation. J. Colloid Interface Sci. 108, 199206.
Lee T. J. and Pielke R. A. (1992) Estimating the soil surface specic humidity. J. Appl. Meteor.
31, 4804.
Lee Y.-N. (1984) Kinetics of some aqueous-phase reactions of peroxyacetyl nitrate. In Gas-
Liquid Chemistry of Natural Waters, Vol.1, BNL 51757, pp. 21/121/7, Brookhaven
National Laboratory, Brookhaven, NY.
Le Henaff P. (1968) M ethodes d etude et propri et es des hydrates, hemiac etals et hemiac etals
deriv es des aldehydes et des c etones. Bull. Soc. Chim. France, 11, 4687700.
Lenschow D. H., Li X. S., Zhu C. J., and Stankov B. B. (1988) The stably stratied boundary
layer over the great planes. Boundary-Layer Meteor. 42, 95121.
Lesins G., Chylek P., and Lohmann U. (2002) Astudy of internal and external mixing scenarios
and its effect on aerosol optical properties and direct radiative forcing. J. Geophys. Res.
107 (D10), doi:10.1029/2001JD000973.
Lettau H. H. (1969) Note on aerodynamic roughness-parameter estimation on the basis of
roughness element description. J. Appl. Meteor. 8, 82832.
Li Z., Williams A. L., and Rood M. J. (1998) Inuence of soluble surfactant properties on
the activation of aerosol particles containing inorganic solute. J. Atmos. Sci. 55, 1859
66.
Liang J. and Jacob D. J. (1997) Effect of aqueous-phase cloud chemistry on tropospheric
ozone. J. Geophys. Res. 102, 59936002.
Liang J. and Jacobson M. Z. (1999) A study of sulfur dioxide oxidation pathways for a range
of liquid water contents, pHs, and temperatures. J. Geophys. Res. 104, 13, 74969.
(2000) Comparison of a 4000-reaction chemical mechanism with the carbon bond IV and
an adjusted carbon bond IV-EX mechanism using SMVGEAR II. Atmos. Environ. 34,
301526.
Lide D. R., ed.-in-chief (2003) CRC Handbook of Chemistry and Physics. Boca Raton, FL,
CRC Press, Inc.
Lilly D. K. (1996) Acomparison of incompressible, anelastic and Boussinesq dynamics. Atmos.
Res. 40, 14351.
Lind J. A. and Kok G. L. (1986) Henrys law determinations for aqueous solutions of hydro-
gen peroxide, methylhydroperoxide, and peroxyacetic acid. J. Geophys. Res. 91, 7889
95.
Lind J. A., Kok G. L., and Lazrus A. L. (1987) Aqueous phase oxidation of sulfur(IV) by
hydrogen peroxide, methylhydroperoxide, and peroxyacetic acid. J. Geophys. Res. 92,
41717.
Lindzen R. S. (1981) Turbulence and stress due to gravity wave and tidal breakdown.
J. Geophys. Res. 86, 970714.
Liou K. N. (1974) Analytic two-stream and four-stream solutions for radiative transfer.
J. Atmos. Sci. 31, 14735.
(2002) An Introduction to Atmospheric Radiation. Amsterdam, Academic Press.
Liousse C., Penner J. E., Chuang C., Walton J. J., Eddleman H., and Cachier H. (1996) A
global three-dimensional model study of carbonaceous aerosols. J. Geophys. Res. 101,
19,41132.
768
References
Liss P. S. and Merlivat L. (1986) Airsea gas exchange rates: Introduction and synthesis. In
The Role of AirSea Exchange in Geochemical Cycling. P. Buat-Menard, ed., Hingham,
MA, D. Reidel Publishing Co., 113127.
List R. J., ed. (1984) Smithsonian Meteorological Tables, 6th edn. Washington, DC, Smithso-
nian Institution Press.
List R. and Gillespie J. R. (1976) Evolution of raindrop spectra with collision-induced breakup.
J. Atmos. Sci. 33, 200713.
Lister J. D., Smit D. J., and Hounslow M. J. (1995) Adjustable discretized population balance
for growth and aggregation. AIChE Journal 41, 591603.
Liu C. H. and Leung D. Y. C. (2001) Turbulence and dispersion studies using a three-
dimensional second-order closure Eulerian model. J. Appl. Meteorol. 40, 92113.
Lord S. J. and Arakawa A. (1980) Interaction of a cumulus cloud ensemble with the large-scale
environment. Part II. J. Atmos. Sci. 37, 267792.
Lorentz H. A. (1906) The absorption and emission of lines of gaseous bodies. In H. A. Lorentz
Collected Papers (The Hague, 19341939) 3, 21538.
Lorenz E. N. (1960) Energy and numerical weather prediction. Tellus 12, 36473.
Louis J.-F. (1979) A parametric model of vertical eddy uxes in the atmosphere. Boundary-
Layer Meteor. 17, 187202.
Lu R. (1994) Development of an integrated air pollution modeling system and simulations of
ozone distributions over the Los Angeles Basin. Ph.D. Thesis, University of California,
Los Angeles.
Lu R. and Turco R. P. (1994) Air pollution transport in a coastal environment. Part I: Two-
dimensional simulations of sea-breeze and mountain effects. J. Atmos. Sci. 51, 2285
308.
Ludlum F. H. (1980) Clouds and Storms. University Park, PA, The Pennsylvania State Univer-
sity Press.
Lurmann F. W., Carter W. P. L., and Coyner L. A. (1987) A Surrogate Species Chemical
Reaction Mechanismfor Urban Scale Air Quality Simulation Models. Volume I: Adaption
of the Mechanism. EPA-600/3-87/014a, U.S. Environmental Protection Agency, Research
Triangle Park, NC.
Lurmann F. W., Main H. H., Knapp K. T., Stockburger L., Rasmussen R. A., and Fung K.
(1992) Analysis of the Ambient VOC Data Collected in the Southern California Air
Quality Study, Final Report to the California Air Resources Board under Contract A832-
130.
Madronich S. (1987) Photodissociation in the atmosphere 1. Actinic ux and the effects of
ground reections and clouds. J. Geophys. Res. 92, 974052.
Madronich, S. and Calvert J. G. (1989) The NCAR Master Mechanism of the Gas-phase
Chemistry-Version 2.0. Rep. NCAR/TN-333+STR, National Center for Atmospheric
Research.
Mahfouf J.-F. and Noilhan J. (1991) Comparative study of various formulations of evaporation
from bare soil using in situ data. J. Appl. Meteor. 30, 135465.
(1996) Inclusion of gravitational drainage in a land surface scheme based on the force-restore
method. J. Appl. Meteor. 35, 98792.
Mahrt L., Heald R. C., Lenschow D. H., Stankov B. B., and Troen I. (1979) An observational
study of the structure of the nocturnal boundary layer. Boundary-Layer Meteor. 17, 247
64.
Makar P. A. (2001) The estimation of organic gas vapour pressure. Atmos. Environ. 35, 961
74.
Makar P. A. and Karpik S. R. (1996) Basis-spline interpolation on the sphere: Applications to
semi-lagrangian advection. Mon. Wea. Rev. 124, 18299.
Makar P. A., Vouchet V. S., and Nenes A. (2003) Inorganic chemistry calculations using
HETV a vectorized solver for the SO
4
2
-NO
3
-NH
4
+
systembased on the ISORROPIA
algorithms. Atmos. Environ. 37, 227994.
769
References
Makar P. A., Moran M. D., Scholtz M. T., and Taylor A. (2003) Speciation of volatile organic
compound emissions for regional air quality modeling of particulate matter and ozone.
J. Geophys. Res. 108 (D2), 4041, doi:10.1029/2001JD000797.
Manabe S. J., Smagorinsky J., and Strickler R. F. (1965) Simulated climatology of a general
circulation model with a hydrological cycle. Mon. Wea. Rev. 93, 76998.
Marbaix P., Gallee H., Brasseur O., and Van Ypersele J.-P. (2003) Lateral boundary conditions
in regional climate models: A detailed study of the relaxation procedure. Mon. Wea. Rev.
131, 46179.
Maricq, M. M., Chase R. E., Podsiadlik D. H., and Vogt R. (1999) Vehicle Exhaust Particle
Size Distributions: A Comparison of Tailpipe and Dilution Tunnel Measurements. SAE
Technical Paper 1999-01-1461, Warrendale, PA, USA.
Marland G., Boden T. A., and Andres R. J. (2003) Global CO
2
emissions from fossil-fuel
burning, cement manufacture, and gas aring: 17512000. In Trends Online: A Com-
pendium of Data on Global Change. Carbon Dioxide Information Analysis Center, Oak
Ridge National Laboratory, U.S. Department of Energy, Oak Ridge, TN, USA.
Marlow W. H. (1981) Size effects in aerosol particle interactions: The van der Waals potential
and collision rates. Surf. Sci. 106, 52937.
Marsh A. R. W. and McElroy W. J. (1985) The dissociation constant and Henrys lawconstant
of HCl in aqueous solution. Atmos. Environ. 19, 107580.
Marshall J. S. and Palmer W. (1948) The distribution of raindrops with size. J. Meteor. 3,
1658.
Marshall S. F., Covert D.S., and Charlson R. J. (1995) Relationship between asymmetry param-
eter and hemispheric backscatter ratio: implications for climate forcing by aerosols. Appl.
Opt. 34, 630611.
Martens C. S., Wesolowski J. J., Hariss R. C., and Kaifer R. (1973) Chlorine loss from Puerto
Rican and San Francisco Bay Area marine aerosols. J. Geophys. Res. 78, 877892.
Martensson E. M., Nilsson E. D., de Leeuw G., Cohen L. H., and Hansson H.-C. (2003)
Laboratory simulation and parameterization of the primary marine aerosol production.
J. Geophys. Res. 108 (D9), 4297, doi:10.1029/2002JD002263.
Marticorena B. and Bergametti G. (1995) Modeling the atmospheric dust cycle: 1. Design of
a soil-derived dust emission scheme. J. Geophys. Res. 100, 1641530.
Marticorena B., Bergametti G., Gillette D., and Belnap J. (1997) Factors controlling threshold
friction velocity in semiarid and arid areas of the United States. J. Geophys. Res. 102,
2327787.
Martin J. J., Wang P. K., and Pruppacher H. R. (1980) A theoretical study of the effect of
electric charges on the efciency with which aerosol particles are collected by ice crystal
plates. J. Colloid Interface Sci. 78, 4456.
Martin L. R. and Hill M. W. (1987a) The iron-catalyzed oxidation of sulfur: Reconciliation
of the literature rates. Atmos. Environ. 21, 148790.
(1987b) The effect of ionic strength on the manganese catalyzed oxidation of sulfur(IV).
Mason B. J. (1971) The Physics of Clouds. Oxford, Clarendon Press.
Matsuno T. (1966) Numerical integrations of the primitive equations by simulated backward
difference scheme. J. Meteor. Soc. Japan 44, 7684.
Mauna Loa Data Center (2001) Data for atmospheric trace gases. http://mloserv.mlo.
hawaii.gov/.
Maxwell J. C. (1890) The Scientic Papers of James Clerk Maxwell, Vol. II. W. D. Niven, ed.,
Cambridge, Cambridge University Press, 63640.
Maxwell Garnett J. C. (1904) Colours in metal glasses and in metallic lms. Philos. Trans.
Roy. Soc. A203, 385420.
McClelland L., Simkin T., Summers M., Nielsen E., and Stein T. C. (eds.) (1989) Global
Volcanism 19751985. Englewood Cliffs, NJ, Prentice-Hall, 655pp.
770
References
McCumber M. C. (1980) A numerical simulation of the inuence of heat and moisture uxes
upon mesoscale circulations, Ph.D. Thesis, University of Virginia, Charlottesville.
McCumber M. C. and Pielke R. A. (1981) Simulation of the effects of surface uxes of heat
and moisture in a mesoscale numerical model. Part I: Soil layer. J. Geophys. Res. 86,
992938.
McElroy M. B., Salawitch R. J., Wofsy S. C., and Logan J. A. (1986) Reduction of Antarctic
ozone due to synergistic interactions of chlorine and bromine. Nature 321, 75962.
McGraw R. and Saunders J. H. (1984) A condensation feedback mechanism for oscillatory
nucleation and growth. Aerosol Sci. Technol. 3, 36780.
McMurry P. H. and Grosjean D. (1985) Photochemical formation of organic aerosols: growth
laws and mechanisms. Atmos. Environ. 19, 144551.
McRae G. J., Goodin W. R., and Seinfeld J. H. (1982) Development of a second-generation
mathematical model for urban air pollution I. Model formulation. Atmos. Environ. 16,
67996.
Meador W. E. and Weaver W. R. (1980) Two-stream approximations to radiative transfer in
planetary atmospheres: Aunied description of existing methods and a newimprovement.
J. Atmos. Sci. 37, 63043.
Mellor G. L. and Yamada T. (1974) A hierarchy of turbulence closure models for planetary
boundary layers. J. Atmos. Sci. 31, 1791806.
(1982) Development of a turbulence closure model for geophysical uid problems. Rev.
Geophys. Space Phys. 20, 85175.
Meng Z. and Seinfeld J. H. (1996) Time scales to achieve atmospheric gasaerosol equilibrium
for volatile species. Atmos. Environ. 30, 2889900.
Meng Z., Seinfeld J. H., Saxena P., and KimY. P. (1995) Atmospheric gasaerosol equilibrium:
IV. Thermodynamics of carbonates. Aerosol. Sci. Technol. 23, 13154.
Meng Z., Dabdub D., and Seinfeld J. H. (1998) Size-resolved and chemically resolved model
of atmospheric aerosol dynamics. J. Geophys. Res. 103, 341935.
Mesinger F. and Arakawa A. (1976) Numerical Methods Used in Atmospheric Models. GARP
Publication Series. No. 17, 1. World Meteorological Organization, 64pp.
Metzger S., Dentener F., Pandis S., and Lelieveld J. (2002) Gas/aerosol partitioning: 1. A
computationally efcient model. J. Geophys. Res. 107 (D16) 10.1029/2001JD001102.
Middleton P. and Brock J. R. (1976) Simulation of aerosol kinetics. J. Colloid Interface Sci.
54, 24964.
Middleton W. E. K. (1952) Vision Through the Atmosphere. Toronto, Canada, University of
Toronto Press.
Mihailovic D. T., Rajkovic B., Lalic B., and Dekic L. (1995) Schemes for parameterizing
evaporation froma non-plant covered surface and their impact on partitioning the surface
energy in landair exchange parameterization. J. Appl. Meteor. 34, 246275.
Millero F. J. (1995) Thermodynamics of the carbon dioxide system in the oceans. Geochim.
Cosmochim. Acta 59, 6617.
Millikan R. A. (1923) The general law of fall of a small spherical body through a gas, and its
bearing upon the nature of molecular reection from surfaces. Phys. Rev. 22, 123.
Mitchell A. R. (1969) Computational Methods in Partial Differential Equations. New York,
John Wiley.
Miyakoda K., Smagorinsky J., Strickler R. F., and Hembree G. D. (1969) Experimental
extended predictions with a nine-level hemispheric model. Mon. Wea. Rev. 97, 176.
Moeng C. H. (1984) A large-eddy simulation model for the study of planetary boundary-layer
turbulence. J. Atmos. Sci. 41, 220216.
Molina L. T. and Molina M. J. (1986) Production of Cl
2
O
2
by the self reaction of the ClO
radical. J. Phys. Chem. 91, 4336.
Molina M. J. and Rowland F. S. (1974) Stratospheric sink for chlorouoromethanes: Chlorine
atom catalysed destruction of ozone. Nature 249, 8102.
771
References
Molinari J. (1982) A method for calculating the effects of deep cumulus convection in numer-
ical models. Mon. Wea. Rev. 11, 152734.
Monahan E. C., Spiel D. E., and Davidson K. L. (1986) A model of marine aerosol generation
via whitecaps and wave disruption. In Oceanic Whitecaps and Their Role in AirSea
Exchange Processes. E. C. Monahan and G. MacNiocaill, eds., Norwell, MA, D. Reidel,
16774.
Monin A. S. and Obukhov A. M. (1954) Basic laws of turbulent mixing in the ground layer
of the atmosphere. Trans. Geophys. Inst. Akad. Nauk USSR 151, 196387.
Monin A. S. and Yaglom (1971) Statistical Fluid Mechanics. Cambridge, MA, MIT Press.
Monteith J. L. and Szeicz G. (1962) Radiative temperature in the heat balance of natural
surfaces. Q. J. Roy. Meteor. Soc. 88, 496507.
Moorthi S. and Suarez M. J. (1992) Relaxed ArakawaSchubert: A parameterization of moist
convection for general circulation models. Mon. Wea. Rev. 120, 97886.
Mordy W. (1959) Computations of the growth by condensation of a population of cloud
droplets. Tellus 11, 1644.
Mountain R. D., Mulholland G. W., and BaumH. (1986) Simulation of aerosol agglomeration
in the free molecular and continuum ow regimes. J. Colloid. Interface Sci. 114, 6781.
Moya M., Pandis S. N., and Jacobson M. Z. (2001) Is the size distribution of urban
aerosols determined by thermodynamic equilibrium? An application to Southern Cali-
fornia. Atmos. Environ. 36, 234965.
Mozurkewich M., McMurry P. H., Gupta A., and Calvert J. G. (1987) Mass accommodation
coefcients for HO
2
radicals on aqueous particles. J. Geophys. Res. 92, 416370.
Mulholland G. W., Samson R. J., Mountain R. D., and Ernst M. H. (1988) Cluster size
distribution for free molecular agglomeration. Energy and Fuels 2, 4816.
Muller H. (1928) Zur allgemeinen Theorie der raschen Koagulation. Die koagulation von
Stabchen- und Blattchen-kolloiden; die Theorie beliebig polydisperser Systeme und der
Stromungskoagulation. Kolloidbeihefte 27, 22350.
Munger W. J., Collett J. Jr., Daube B. C., and Hoffmann M. R. (1989) Carboxylic acids and
carbonyl compounds in southern California clouds and fogs. Tellus 41b, 23042.
Mylonas D. T., Allen D. T., Ehrman S. H., and Pratsinis S. E. (1991) The sources and size
distributions of organonitrates in Los Angeles aerosols. Atmos. Environ. 25A, 285561.
Nair R. D., Scroggs J. S., and Semazzi F. H. M. (2002) Efcient conservative global transport
schemes for climate and atmospheric chemistry models. Mon. Wea. Rev. 130, 205973.
Nakicenovic N. and Swart, R., eds. (2000) Emissions Scenarios. A Special Report of the
Intergovernmental Panel on Climate Change. Cambridge, Cambridge University Press.
Napari I., Noppel M., Vehkamaki H., and Kulmala M. (2002) Parameterization of ternary
nucleation rates for H
2
SO
4
-NH
3
-H
2
O vapors. J. Geophys. Res. 107 (D19), 4381,
doi:10.1029/2002JD002132.
National Oceanic and Atmospheric Administration (NOAA) (1976) U.S. Standard Atmo-
sphere. Washington, DC.
Naumann K.-H. (2003) COSIMA-A computer program simulating the dynamics of fractal
aerosols. J. Aerosol Sci. 34, 137197.
Nautical Almanac Ofce (NAO) and Her Majestys Nautical Almanac Ofce (1993) Astro-
nomical Almanac. Washington, DC, U.S. Government Printing Ofce.
Nebeker F. (1995) Calculating the Weather. San Diego, Academic Press, Inc.
Nenes A., Pandis S. N., and Pilinis C. (1998) ISORROPIA: A newthermodynamic equilibrium
model for multiphase multicomponent inorganic aerosols. Aquat. Geochem. 4, 12352.
(1999) Continued development and testing of a new thermodynamic aerosol module for
urban and regional air quality. Atmos. Environ. 33, 15531560.
Nesbitt F. L., Monks P. S., Wayne W. A., Stief L. J., and Touni R. (1995) The reaction of O(
3
P)
+ HOBr: Temperature dependence of the rate constant and importance of the reaction as
an HOBr loss process. Geophys. Res. Lett. 22, 82730.
772
References
Nguyen K. and Dabdub D. (2001) Two-level time-marching scheme using splines for solving
the advection equation. Atmos. Environ. 35, 162737.
(2002) Semi-Lagrangian ux scheme for the solution of the aerosol condensation/
evaporation equation. Aerosol Sci. Technol. 36, 407418.
Nicolet M. (1989) Solar spectral irradiances with their diversity between 120 and 900 nm.
Planet. Space Sci. 37, 124989.
Noilhan J. and Planton S. (1989) A simple parameterization of land surface processes for
meteorological models. Mon. Wea. Rev. 117, 53649.
Noll K. E., Fang K. Y. P., and Khalili E. (1990) Characterization of atmospheric coarse particles
in the Los Angeles Basin. Aerosol Sci. Technol. 12, 2838.
Odum J. R., Hoffmann T., Bowman F., Collins T., Flagan R. C., and Seinfeld J. H. (1996)
Gas-particle partitioning and secondary organic aerosol yields. Environ. Sci. Technol. 30,
25805.
Ogura Y. and Phillips N. A. (1962) Scale analysis of deep and shallow convection in the
atmosphere. J. Atmos. Sci. 19, 1739.
Okada K. and Hitzenberger R. (2001) Mixing properties of individual submicrometer particles
in Vienna. Atmos. Environ. 35, 561728.
Oke T. R. (1978) Boundary Layer Climates. London, Methuen.
Oke T. R., Spronken-Smith R. A., Jauregui E., and Grimmond C. S. B. (1999) The energy
balance of central Mexico City during the dry season. Atmos. Environ. 33, 391930.
Okuyama K., Kousaka Y., and Hayashi K. (1984) Change in size distribution of ultrane
aerosol particles undergoing Brownian coagulation. J. Colloid Interface Sci. 101, 98
109.
Olscamp P. J., translator (1965) Discourse on Method, Optics, Geometry, and Meteorology
by Ren e Descartes. Indianapolis, Bobbs-Merrill Company, Inc.
Ooyama V. K. (1971) A theory on parameterization of cumulus convection. J. Meteor. Soc.
Japan 49, 74456.
Orszag S. A. (1970) Transform method for calculation of vector coupled sums: Application to
the spectral form of the vorticity equation. J. Atmos. Sci. 27, 8905.
(1971) Numerical simulation of incompressible ows within simple boundaries. I. Galerkin
(spectral) representations. Stud. Appl. Math. 50, 293326.
Orville H. D. and Kopp F. J. (1977). Numerical simulations of the history of a hailstorm.
J. Atmos. Sci. 34, 1596618.
Osborne N. S., Stimson H. F., and Ginnings D. C. (1939) Measurements of heat capacity and
heat of vaporization of water in the range of 0 degrees to 100 degrees celsius. J. Res. Nat.
Bur. Stand. 23, 197260.
Pandis S. N. and Seinfeld J. H. (1989) Sensitivity analysis of a chemical mechanismfor aqueous-
phase atmospheric chemistry. J. Geophys. Res. 94, 110526.
Pandis S. N., Harley R. A., Cass G. R., and Seinfeld J. H. (1992) Secondary organic aerosol
formation and transport. Atmos. Environ. 26A, 226982.
Pandis S. N., Russell L. M., and Seinfeld J. H. (1994) The relationship between DMS ux and
CCN concentration in remote marine regions. J. Geophys. Res. 99, 1694557.
Park J.-Y. and Lee Y.-N. (1987) Aqueous solubility and hydrolysis kinetics of peroxynitric
acid. Paper presented at 193rd Meeting, American Chemical Society, Denver, CO, April
510.
Parker V. B. (1965) Thermal Properties of Aqueous Uni-univalent Electrolytes. National Stan-
dard Reference Data Series NBS 2. U.S. Government Printing Ofce, Washington, DC.
Parkinson C. L. and Washington W. M. (1979) A large-scale numerical model for sea ice.
Pasquill F. (1962) Atmospheric Diffusion. London, Van Nostrand.
Paulson S. E. and Seinfeld J. H. (1992) Development and evaluation of a photooxidation
mechanism for isoprene. J. Geophys. Res. 97, 20,70315.
773
References
Peng C., Chan M. N., and Chan C. K. (2001) The hygroscopic properties of dicarboxylic and
multifunctional acids: Measurements and UNIFAC predictions. Environ. Sci. Technol.
35, 4495501.
Pepper D. W., Kern C. D., and Long P. E. Jr. (1979) Modeling the dispersion of atmospheric
pollution using cubic splines and chapeau functions. Atmos. Environ. 13, 22337.
Perrin D. D. (1982) Ionization Constants of Inorganic Acids and Bases in Aqueous Solution,
2nd edn. New York, Pergamon.
Perron G., Roux A., and Desnoyers J. E. (1981) Heat capacities and volumes of NaCl, MgCl
2
,
CaCl
2
, and NiCl
2
up to 6 molal in water. Can. J. Chem. 59, 304954.
Petersen R. L. (1997) A wind tunnel evaluation of methods for estimating surface roughness
length at industrial facilities. Atmos. Environ. 31, 4557.
Philip J. R. (1957) Evaporation, and moisture and heat elds in the soil. J. Meteor. 14, 35466.
Phillips N. A. (1957) A coordinate system having some special advantages for numerical
forecasting. J. Meteor. 14, 1845.
Pielke R. A. (1984) Mesoscale Meteorological Modeling. San Diego, Academic Press, Inc.
Pilinis C. and Seinfeld J. H. (1987) Continued development of a general equilibrium model
for inorganic multicomponent atmospheric aerosols. Atmos. Environ. 21, 245366.
(1988) Development and evaluation of an eulerian photochemical gas-aerosol model.
Pilinis C., Capaldo K. P., Nenes A., and Pandis S. N. (2000) MADM-A new multicomponent
aerosol dynamics model. Aerosol Sci. Technol. 32, 482502.
Pinto J. P., Turco R. P., and Toon O. B. (1989) Self-limiting physical and chemical effects in
volcanic eruption clouds. J. Geophys. Res. 94, 11,165.
Pitter R. L. and Pruppacher H. R. (1973) A wind tunnel investigation of freezing of small
water drops falling at terminal velocity in air. Q. J. Roy. Meteor. Soc. 99, 54050.
Pitzer K. S. (1991) Ion interaction approach: Theory and data correlation. In Activity Coef-
cients in Electrolyte Solutions, 2nd edn. K. S. Pitzer, ed., Boca Raton, FL, CRC Press,
75153.
Pitzer K. S. and Mayorga G. (1973) Thermodynamics of electrolytes II. Activity and osmotic
coefcients for strong electrolytes with one or both ions univalent. J. Phys. Chem. 77,
23008.
Pollack J. B. and Cuzzi J. N. (1980) Scattering by nonspherical particles of size comparable to
a wavelength: A new semi-empirical theory and its application to tropospheric aerosols.
J. Atmos. Sci. 37, 86881.
Pooley F. D. and Mille M. (1999) Composition of air pollution particles. In Air Pollution and
Health. S. T. Holgate, J. M. Samet, H. S. Koren, and R. L. Maynard, eds., San Diego,
Academic Press, 61934.
Potter J. F. (1970) The delta-function approximation in radiative transfer theory. J. Atmos.
Sci. 27, 9439.
Prather M. J. (1986) Numerical advection by conservation of second-order moments. J. Geo-
phys. Res. 91, 667181.
Pratsinis S. E. (1988) Simultaneous nucleation, condensation, and coagulation in aerosol reac-
tors. J. Colloid Interface Sci. 124, 41627.
Press W. H., Flannery B. P., Teukolsky, S. A., and Vetterling W. T. (1992). Numerical Recipes:
The Art of Scientic Computing. Cambridge, Cambridge University Press.
Price C., Penner J., and Prather M. (1997) NOx from lightning 1. Global distribution based
on lightning physics. J. Geophys. Res. 102, 592941.
Price G. V. and MacPherson A. K. (1973) A numerical weather forecasting method using cubic
splines on a variable mesh. J. Appl. Meteor. 12, 110213.
Pruppacher H. R. and Klett J. D. (1997) Microphysics of Clouds and Precipitation, 2nd rev.
and enl. edn., Dordrecht, Kluwer Academic Publishers.
Pruppacher H. R. and Rasmussen R. (1979) A wind tunnel investigation of the rate of evapo-
ration of large water drops falling at terminal velocity in air. J. Atmos. Sci. 36, 125560.
774
References
Purnell D. K. (1976) Solution of the advection equation by upstreaminterpolation with a cubic
spline. Mon. Wea. Rev. 104, 428.
Rahmes T. F., Omar A. H., and Wuebbles D. J. (1998) Atmospheric distributions of soot
particles by current and future aircraft eets and resulting radiative forcing on climate.
J. Geophys. Res. 103, 31,65767.
Rao N. P. and McMurry P. H. (1989) Nucleation and growth of aerosol in chemically reacting
systems. Aerosol Sci. Technol. 11, 12033.
Raoult F.-M. (1887) General law of the vapor pressure of solvents. Comptes Rendus 104,
14303.
Rasch P. J. (1994) Conservative shape-preserving two-dimensional transport on a spherical
grid. Mon. Wea. Rev. 122, 133750.
Rasmussen R. and Pruppacher H. R. (1982) Awind tunnel and theoretical study of the melting
behavior of atmospheric ice particles. I: A wind tunnel study of frozen drops of radius
-500 mm. J. Atmos. Sci. 39, 1528.
Rasmussen R., Levizzani M. V., and Pruppacher H. R. (1984) A wind tunnel and theoretical
study of the melting behavior of atmospheric ice particles. II: Atheoretical study for frozen
drops of radius -500 mm. J. Atmos. Sci. 41, 37480.
Reid J. S. and Hobbs P. V. (1998) Physical and optical properties of young smoke from indi-
vidual biomass res in Brazil. J. Geophys. Res. 103, 32,01330.
Reid, J. S., Hobbs P. V., Ferek R. J., et al. (1998) Physical, chemical, and optical properties of
regional hazes dominated by smoke in Brazil. J. Geophys. Res. 103, 32,05980.
Reisin T., Levin Z., and Tzivion S. (1996) Rain production in convective clouds as simulated
in an axisymmetric model with detailed microphysics. Part I: Description of the model.
J. Atmos. Sci. 53, 497519.
Reynolds S. D., Roth P. M., and Seinfeld J. H. (1973) Mathematical modeling of photochemical
air pollution I: Formulation of the model. Atmos. Environ. 7, 103361.
Richardson L. F. (1922) Weather Prediction by Numerical Process. Cambridge, Cambridge
University Press, reprinted 1965, 236pp.
Robert A. (1982) A semi-Lagrangian and semi-implicit numerical integration scheme for the
primitive meteorological equations. Japan Meteor. Soc. 60, 31925.
Robinson R. A. and Stokes R. H. (1955) Electrolyte Solutions. New York, Academic Press.
Rogak S. N. and Flagan R. C. (1992) Coagulation of aerosol agglomerates in the transition
regime. J. Colloid Interface Sci. 151, 20324.
Rogers R. R. and Yau M. K. (1989) A Short Course in Cloud Physics. Oxford, Pergamon
Press.
RosenbaumJ. S. (1976) Conservation properties of numerical integration methods for systems
of ordinary differential equations. J. Comp. Phys. 20, 25967.
Ross A. B. and Neta P. (1979) Rate Constants for Reactions of Inorganic Radicals in Aqueous
Solutions. NSRDS-NBS 65. National Bureau of Standards, U.S. Department of Com-
merce, Washington, DC.
Rossby C. and collaborators (1939) Relation between variations in the intensity of the zonal
circulation of the atmosphere and the displacements of the semi-permanent centers of
action. J. Marine Res. 2, 3855.
Rothman L. S., et al. (2003) The HITRAN molecular spectroscopic database: Edition of 2000
including updates of 2001. J. Quant. Spectrosc. Radiat. Transfer 82, 544.
Roux A., Musbally G. M., Perron G., et al. (1978) Apparent molal heat capacities and volumes
of aqueous electrolytes at 25

C: NaClO
3
, NaClO
4
, NaNO
3
, NaBrO
3
, NaIO
3
, KClO
3
,
KBrO
3
, KIO
3
, NH
4
NO
3
, NH
4
Cl, and NH
4
ClO
4
. Can. J. Chem. 56, 248.
Russell A. G., Winner D. A., Harley R. A., McCue K. F., and Cass G. R. (1993) Mathematical
modeling and control of the dry deposition ux of nitrogen-containing air pollutants.
Environ. Sci. Technol. 27, 277282.
Russell L. M., Pandis S. N., and Seinfeld J. H. (1994) Aerosol production and growth in the
marine boundary layer. J. Geophys. Res. 99, 20,98921,003.
775
References
Saffman P. G. and Turner J. S. (1956) On the collision of drops in turbulent clouds. J. Fluid
Mech. 1, 1630.
Sander R., Lelieveld J., and Crutzen P. J. (1995) Modelling of nighttime nitrogen and sulfur
chemistry in size resolved droplets of an orographic cloud. J. Atmos. Chem. 20, 89116.
Sandu A. (2001) Positive numerical integration methods for chemical kinetic systems. J. Comp.
Phys. 170, 589602.
(2002) A NewtonCotes quadrature approach for solving the aerosol coagulation equation.
Sandu A., Verwer J. G., van Loon M., et al. (1997) Benchmarking stiff ODE solvers for
atmospheric chemistry problems I: Implicit versus explicit. Atmos. Environ. 31, 3151
66.
San Jose R., Casanova J. L., Viloria R. E., and Casanova J. (1985) Evaluation of the turbu-
lent parameters of the unstable surface boundary layer outside Busingers range. Atmos.
Environ. 19, 155561.
Saunders S. M., Jenkin M. E., Derwent R. G., and Pilling M. J. (2003) Protocol for the devel-
opment of the Mater Chemical Mechanism, MCM v3 (Part A): tropospheric degradation
of non-aromatic volatile organic compounds. Atmos. Chem. Phys. 3, 16180.
Saxena P., Hudischewskyj A. B., Seigneur C., and Seinfeld J. H. (1986) A comparative study of
equilibrium approaches to the chemical characterization of secondary aerosols. Atmos.
Environ. 20, 147183.
Saxena P., Mueller P. K., and Hildemann L. M. (1993) Sources and chemistry of chloride
in the troposphere: A review. In Managing Hazardous Air Pollutants: State of the Art.
W. Chow and K. K. Connor, eds., Boca Raton, FL, Lewis Publishers, 17390.
Schmidt K. H. (1972) Electrical conductivity techniques for studying the kinetics of radiation-
induced chemical reactions in aqueous solutions. Int. J. Radiat. Phys. Chem. 4, 43968.
Schmidt-Ott A. and Burtscher H. (1982) The effect of van der Waals forces on aerosol coag-
ulation. J. Colloid Interface Sci. 89, 3537.
Schnaiter M., Horvath H., Mohler O., Naumann K.-H., Saathoff H., and Schock O. W. (2003)
UV-VIS-NIR spectral optical properties of soot and soot-containing aerosols. J. Aerosol
Sci. 34, 142144.
Schneider W., Moortgat G. K., Tyndall G. S., and Burrows J. P. (1987) Absorption cross-
sections of NO
2
in the UV and visible region (200700 nm) at 298 K. J. Photochem.
Photobiol, A: Chem. 40, 195217.
Scholes G. and Willson R. L. (1967) -radiolysis of aqueous thymine solutions. Determination
of relative reaction rates of OH radicals. Trans. Faraday Soc. 63, 298293.
Schroeder W. H., Dobson M., Kane D. M., and Johnson N. D. (1987) Toxic trace elements
associated with airborne particulate matter: a review. J. Air Pollut. Control Assoc. 37,
126785.
Schwartz S. E. (1984) Gas- and aqueous-phase chemistry of HO
2
in liquid water clouds.
J. Geophys. Res. 89, 11,58998.
(1986) Mass-transport considerations pertinent to aqueous phase reactions of gases in
liquid-water clouds. In Chemistry of Multiphase Atmospheric Systems, NATOASI Series,
Vol. G6. W. Jaeschke, ed., Berlin, Springer-Verlag, 41571.
Schwartz S. E. and White W. H. (1981) Solubility equilibria of the nitrogen oxides and oxyacids
in aqueous solution. Adv. Environ. Sci. Eng. 4, 145.
Seaman N. L., Ludwig F. L., Donall E. G., Warner T. T., and Bhumralkar C. M. (1989)
Numerical studies of urban planetary boundary-layer structure under realistic synoptic
conditions. J. Appl. Meteor. 28, 76081.
Seery D. J. and Britton D. (1964) The continuous absorption spectra of chlorine, bromine,
bromine chloride, iodine chloride, and iodine bromide. J. Phys. Chem. 68, 22636.
Sehested K., Rasmussen O. L., and Fricke H. (1968) Rate constants of OH with HO
2
, O
2
,
and H
2
O
2
+
from hydrogen peroxide formation in pulse-irradiated oxygenated water. J.
Phys. Chem. 72, 62631.
776
References
Sehmel G. A. (1980) Particle and gas dry deposition: A review. Atmos. Environ. 14, 983
1011.
Seinfeld J. H. and Pandis S. N. (1998) Atmospheric Chemistry and Physics. New York, Wiley-
Interscience.
Sellers W. D. (1965) Physical Climatology. Chicago, University of Chicago Press, 272pp.
Sellers, P. J., Los S. O., Tucker C. J., et al. (1996) Arevised land surface parameterization (SiB2)
for atmospheric GCMs. Part II: The generation of global elds of terrestrial biophysical
parameters from satellite data. J. Clim. 9, 70637.
Shao Y. (2001) A model for mineral dust emission. J. Geophys. Res. 106, 2023954.
Shao Y., Raupach M. R., and Leys J. F. (1996) A model for predicting Aeolian sand drift and
dust entrainment on scales from paddock to region. Aust. J. Soil Res. 34, 30942.
Shen T.-L., Wooldridge P. J., and Molina M. J. (1995) Stratospheric pollution and ozone
depletion. In Composition, Chemistry, and Climate of the Atmosphere. H. B. Singh, ed.,
New York, Van Nostrand Reinhold.
Sheridan P. J., Brock C. A., and Wilson J. C. (1994) Aerosol particles in the upper troposphere
and lower stratosphere: Elemental composition and morphology of individual particles
in northern midlatitudes. Geophys. Res. Lett. 21, 258790.
Sherman A. H. and Hindmarsh A. C. (1980) GEARS: APackage for the Solution of Sparse, Stiff
Ordinary Differential Equations. Report UCRL-84102, Lawrence Livermore Laboratory.
Shimazaki T. and Laird A. R. (1970) Amodel calculation of the diurnal variation in minor neu-
tral constituents in the mesosphere and lower thermosphere including transport effects.
J. Geophys. Res. 75, 322135.
Shir C. C. and Bornstein R. D. (1976) Eddy exchange coefcients in numerical models of the
planetary boundary layer. Boundary-Layer Meteor. 11, 17185.
Shuttleworth W. J. (1989) Micrometeorology of temperate and tropical forest. Phil. Trans.
Roy. Soc. London B324, 299334.
Siegel R. and Howell J. R. (1992) Thermal Radiation Heat Transfer. Washington, DC, Taylor
and Francis.
Singh H. B. (1995) Halogens in the atmospheric environment. In Composition, Chemistry,
and Climate of the Atmosphere. H. B. Singh, ed., New York, Van Nostrand Reinhold.
Singh H. B., Viezee W., and Salas L. J. (1988) Measurements of selected C
2
-C
5
hydrocarbons
in the troposphere: Latitudinal, vertical, and temporal variations. J. Geophys. Res. 93,
15,86178.
Singh H. B., Kanakidou M., Crutzen P. J., and Jacob D. J. (1995) High concentrations and
photochemical fate of oxygenated hydrocarbons in the global troposphere. Nature 378,
504.
Singh H. B., Herlth D., Kolyer R., et al. (1996) Reactive nitrogen and ozone over the western
Pacic: Distributions, partitioning, and sources. J. Geophys. Res. 101, 1793808.
Skamarock W. C. and Klemp J. B. (1992) The stability of time-split numerical methods for the
hydrostatic and the nonhydrostatic elastic equations. Mon. Wea. Rev. 120, 210927.
Skamarock W. C., Dye J. E., Defer E., Barth M. C., Stith J. L., and Ridley B. A. (2003)
Observational- and modeling-based budget of lightning-produced NOx in a continental
thunderstorm. J. Geophys. Res. 108 (D10), 4305, doi:10.1029/2002JD002163.
Slinn W. G. N., Hasse L., Hicks B. B., et al. (1978) Some aspects of the transfer of atmospheric
trace constituents past the airsea interface. Atmos. Environ. 12, 205587.
Smith M. H. and Harrison N. M. (1998) The sea spray generation function. J. Aerosol Sci.
29, Suppl. 1, S189S190.
Smith R. M. and Martell A. E. (1976) Critical Stability Constants, Vol. 4: Inorganic Complexes.
New York, Plenum.
Smolarkiewicz P. K. (1983) A simple positive denite advection scheme with small implicit
diffusion. Mon. Wea. Rev. 111, 47986.
Smoluchowski M. V. (1918) Versuch einer mathematischen Theorie der Koagulationskinetik
kolloider L osungen. Z. Phys. Chem. 92, 12968.
777
References
Snider J. R. and Dawson G. A. (1985) Tropospheric light alcohols, carbonyls, and acetonitrile:
Concentrations in the southwestern United States and Henrys law data. J. Geophys. Res.
90, 3797805.
Snyder J. P. (1987) Map Projections A Working Manual. U.S. Geological Survey professional
paper 1395, U.S. Government Printing Ofce, Washington.
Sokolik I., Andronova A., and Johnson C. (1993) Complex refractive index of atmospheric
dust aerosols. Atmos. Environ. 27A, 2495502.
Solomon, S., Garcia R. R., Rowland F. S., and Wuebbles D. J. (1986) On the depletion of
Antarctic ozone. Nature 321, 7557.
Sommer L. (1989) Analytical Absorption Spectrophotometry in the Visible and Ultraviolet.
Amsterdam, Elsevier.
Spencer J. W. (1971) Fourier series representation of the position of the Sun. Search 2, 172.
StamD. M., de Haan J. F., Hovenier J. W., and Stammes P. (2000) Afast method for simulating
observations of polarized light emerging from the atmosphere applied to the oxygen-A
band. J. Quant. Spectrosc. Radiat. Transfer 64, 13149.
Staniforth A. and Cote J. (1991) Semi-Lagrangian integration schemes for atmospheric
models a review. Mon. Wea. Rev. 119, 220623.
Steiner D, Burtchnew H., and Grass H. (1992) Structure and disposition of particles from a
spark ignition engine. Atmos. Environ. 26, 9971003.
Stelson A. W., Bassett M. E., and Seinfeld J. H. (1984) Thermodynamic equilibrium properties
of aqueous solutions of nitrate, sulfate and ammonium. In Chemistry of Particles, Fogs
and Rain. J. L. Durham, ed., Ann Arbor, MI, Ann Arbor Publication, 152.
Stephens E. R., Scott W. E., Hanst P. L., and Doerr R. C. (1956) Recent developments in the
study of the organic chemistry of the atmosphere. J. Air Pollut. Contr. Assoc. 6, 159
65.
Stockwell W. R. (1986) A homogeneous gas-phase mechanism for use in a regional acid depo-
sition model. Atmos. Environ. 20, 161532.
Stockwell W. R. (1995) On the HO
2
+HO
2
reaction: Its misapplication in atmospheric chem-
istry models. J. Geophys. Res. 100, 11,6958.
Stoer J. and Bulirsch R. (1980) Introduction to Numerical Analysis. New York, Springer-
Verlag.
Stokes R. H. and Robinson R. A. (1966) Interactions in aqueous nonelectrolyte solutions.
I. Solutesolvent equilibria. J. Phys. Chem. 70, 212630.
Stommel H. (1947) Entrainment of air into a cumulus cloud. Part I. J. Appl. Meteor. 4, 914.
Streets D. G. and Waldhoff S. T. (1998) Biofuel use in Asia and acidifying emissions, Energy
23, 102942.
(1999) Greenhouse-gas emissions from biofuel combustion in Asia. Energy 24, 84155.
Strom J., Okada K., and Heintzenber J. (1992) On the state of mixing of particles due to
Brownian coagulation. J. Aerosol Sci. 23 46780.
Stuart A. L. (2002) Volatile chemical partitioning during cloud hydrometeor freezing and its
effects on tropospheric chemical distributions. Ph.D. Thesis, Stanford University.
Stull R. B. (1988) An Introduction to Boundary Layer Meteorology. Dordrecht, Kluwer
Academic Publishers.
StummW. and Morgan J. J. (1981) Aquatic Chemistry. NewYork, Wiley Interscience, 780pp.
Suck S. H. and Brock J. R. (1979) Evolution of atmospheric aerosol particle size distributions
via Brownian coagulation: Numerical simulation. J. Aerosol Sci. 10, 58190.
Sukhatme S. P. and Saikhedkar N. (1969) Heat capacities of glycerol-water mixtures and
aqueous solutions of ammonium sulfate, ammonium nitrate and strontium nitrate. Ind.
J. Technol. 7, 14.
Sun Q. and Wexler A. S. (1998) Modeling urban and regional aerosols condensation and
evaporation near acid neutrality. Atmos. Environ. 32, 352731.
Tabazadeh A. and Turco R. P. (1993a) Stratospheric chlorine injection by volcanic eruptions:
HCl scavenging and implications for ozone. Science 260, 10826.
778
References
(1993b) A model for heterogeneous chemical processes on the surfaces of ice and nitric acid
trihydrate particles. J. Geophys. Res. 98, 12,72740.
Tabazadeh A., Turco R. P., Drdla K., and Jacobson M. Z. (1994) A study of Type I polar
stratospheric cloud formation. Geophys. Res. Lett. 21, 161922.
Tabazadeh A., Djikaev Y. S., and Reiss H. (2002) Surface crystallization of supercooled water
in clouds. Proc. Nat. Acad. Sci. 99 15,8738.
Tang I. N. (1996) Chemical and size effects of hygroscopic aerosols on light scattering coef-
cients. J. Geophys. Res. 101, 19,24550.
(1997) Thermodynamic and optical properties of mixed-salt aerosols of atmospheric impor-
tance. J. Geophys. Res. 102, 188393.
Tang I. N. and Munkelwitz H. R. (1993) Composition and temperature dependence of the
deliquescence properties of hygroscopic aerosols. Atmos. Environ. 27A, 46773.
(1994) Water activities, densities, and refractive indices of aqueous sulfates and sodium
nitrate droplets of atmospheric importance. J. Geophys. Res. 99, 18,8018.
Tang I. N., Wong W. T., and Munkelwitz H. R. (1981) The relative importance of atmospheric
sulfates and nitrates in visibility reduction. Atmos. Environ. 15, 246371.
Tanguay M., Robert A., and Laprise R. (1990) A semiimplicit semiLagrangian fully-
compressible regional forecast model. Mon. Wea. Rev. 118, 197080.
Tao Y. and McMurry P. H. (1989) Vapor pressures and surface free energies of C
14
-C
19
mono-
carboxylic acids and C
5
-dicarboxylic and C
6
-dicarboxylic acids. Environ. Sci. Technol.
25, 178893.
Tapp M. C. and White P. W. (1976) A nonhydrostatic mesoscale model. Quart. J. Roy. Meteor.
Soc. 102, 27796.
Tegen I., Lacis A. A., and Fung I. (1996) The inuence on climate forcing of mineral aerosols
from disturbed soils. Nature 380, 41922.
Terry D. A., McGrawR., and Rangel R. H. (2001) Method of moments solutions for a laminar
ow aerosol reactor model. Aerosol Sci. Technol. 34, 35362.
Tesche T. W. (1988) Accuracy of ozone air quality models. J. Environ. Eng. 114, 73952.
Thekaekara M. P. (1974) Extraterrestrial solar spectrum, 30006100 at 1- intervals. Appl.
Opt. 13, 51822.
Thompson N., Barrie N., and Ayles M. (1981) The meteorological ofce rainfall and evapo-
ration calculation system: MORECS. Hydrol. Memo. 45, 169.
Thuburn J. (1996) Multidimensional ux-limited advection schemes. J. Comp. Phys. 123,
7483.
(1997) TVD schemes, positive schemes, and the universal limiter. Mon. Wea. Rev. 125,
19903.
Tiedtke M. (1989) A comprehensive mass ux scheme for cumulus parameterization in large-
scale models. Mon. Wea. Rev. 117, 1779800.
Tjernstrom M. (1993) Turbulence length scales in stably stratied free shear ow analyzed
from slant aircraft proles. J. Appl. Meteor. 32, 94863.
Toon O. B. and Ackerman T. P. (1981) Algorithms for the calculation of scattering by stratied
spheres. Appl. Opt. 20, 365760.
Toon O. B., Hamill P., Turco R. P., and Pinto J. (1986) Condensation of HNO
3
and HCl in
the winter polar stratospheres. Geophys. Res. Lett. Nov. Supp. 13, 12847.
Toon O. B., Turco R. P., Westphal D., Malone R., and Liu M. S. (1988) A multidimensional
model for aerosols: Description of computational analogs. J. Atmos. Sci. 45, 212343.
Toon O. B., McKay C. P., and Ackerman T. P. (1989a) Rapid calculation of radiative heating
rates and photodissociation rates in inhomogeneous multiple scattering atmospheres. J.
Geophys. Res. 94, 16,287301.
Toon O. B., Turco R. P., Jordan J., Goodman J., and Ferry G. (1989b) Physical processes in
polar stratospheric ice clouds. J. Geophys. Res. 94, 11,35980.
Trautmann T. and Wanner C. (1999) A fast and efcient modied sectional method for simu-
lating multicomponent collisional kinetics. Atmos. Environ. 33, 163140.
779
References
Tremback C. J., Powell J., Cotton W. R., and Pielke R. A. (1987) The forward-in-time upstream
advection scheme: Extension to higher orders. Mon. Wea. Rev. 115, 54055.
Troe J. (1979) Predictive possibilities of unimolecular rate theory. J. Phys. Chem. 83, 114
26.
Tsang T. H. and Brock J. R. (1982) Aerosol coagulation in the plume from a cross-wind line
source. Atmos. Environ. 16, 222935.
(1986) Simulation of condensation aerosol growth by condensation and evaporation.
Tsang, T. H. and Huang L. K. (1990) On a PetrovGalerkin nite element method for evapo-
ration of polydisperse aerosols. Aerosol Sci. Technol. 12, 57897.
Tsang T. H. and Korgaonkar N. (1987) Effect of evaporation on the extinction coefcient of
an aerosol cloud. Aerosol Sci. Technol. 7, 31728.
Turco R. P. and Whitten R. C. (1974) A comparison of several computational techniques for
solving some common aeronomic problems. J. Geophys. Res. 79, 317985.
Turco R. P., Hamill P., Toon O. B., Whitten R. C., and Kiang C. S. (1979) The NASA-Ames
Research Center Stratospheric Aerosol Model: I. Physical Processes and Computational
Analogs. NASA Technical Publication (TP) 1362, iii94.
Turco R. P., Toon O. B., Whitten R. C., Keesee R. G., and Hollenbach D. (1982) Noctilucent
clouds: Simulation studies of their genesis, properties and global inuence. Planet. Space
Sci. 30, 114781.
Turco R. P., Toon O. B., and Hamill P. (1989) Heterogeneous physiochemistry of the polar
ozone hole. J. Geophys. Res. 94, 16,493510.
Twohy C. H., Clarke A. D., Warren S. G., Radke L. F., and Charlson R. J. (1989)
Light-absorbing material extracted from cloud droplets and its effect on cloud albedo.
J. Geophys. Res. 94, 862331.
Tyndall G. S. and Ravishankara A. R. (1991) Atmospheric oxidation of reduced sulfur species.
Int. J. Chem. Kinet. 23, 483527.
Tzivion S., Feingold G., and Levin Z. (1987) An efcient numerical solution to the stochastic
collection equation. J. Atmos. Sci. 44, 313949.
U.S. Department of the Army (1958) Universal Transverse Mercator Grid. Tables for Transfor-
mation of Coordinates fromGrid to Geographic; Clarke 1866 Spheroid. U.S. Government
Printing Ofce, Washington, DC.
U.S. Environmental Protection Agency (USEPA) (1978) Air Quality Criteria for Ozone and
Other Photochemical Oxidants. Report No. EPA-600/8-78-004.
Vali G. (1971) Quantitative evaluation of experimental results on the heterogeneous freezing
nucleation of supercooled liquids. J. Atmos. Sci. 28, 4029.
van de Hulst H. C. (1957) Light Scattering by Small Particles. New York, John Wiley and
Sons, Inc.
Vanderzee C. E., Waugh D. H., and Haas N. C. (1980) Enthalpies of dilution and relative
apparent molar enthalpies of aqueous ammoniumnitrate. The case of a weakly hydrolysed
(dissociated) salt. J Chem. Thermodynam. 12, 215.
van Dingenen R. and Raes F. (1993) Ternary nucleation of methane sulphonic acid, sulphuric
acid and water vapour. J. Aerosol Sci. 24, 117.
van Doren J. M, Watson L. R., Davidovits P., Worsnop D. R., Zahniser S., and Kolb C. E.
(1990) Temperature dependence of the uptake coefcients of HNO
3
, HCl, and N
2
O
5
by
water droplets. J. Phys. Chem. 94, 325669.
van Genuchten M. T. (1980) A closed-form equation for predicting the hydraulic conductivity
of unsaturated soils. Ann. Geophys. 3, 61528.
van Weele M. and Duynkerke P. G. (1993) Effects of clouds on the photodissociation of NO
2
:
Observation and modelling. J. Atmos. Chem. 16, 23155.
van Zandt T. E. and Fritts D. C. (1989) A theory of enhanced saturation of the gravity wave
spectrum due to increases in atmospheric stability. Pure Appl. Geophys. Pageoph. 130,
399420.
780
References
Varoglu E. and Finn W. D. L. (1980) Finite elements incorporating characteristics for one-
dimensional diffusion-convection equation. J. Comp. Phys. 34, 37189.
Vehkamaki H., Kulmala M., Napari I., et al. (2002) An improved parameterization for sulfuric
acid-water nucleation rates for tropospheric and stratospheric conditions. J. Geophys.
Res. 107 (D22), 4622, doi:10.1029/2002JD002184.
Venkataraman C. and Friedlander S. K. (1994) Size distributions of polycyclic aromatic hydro-
carbons and elemental carbon. 2. Ambient measurements and effects of atmospheric pro-
cesses. Environ. Sci. Technol. 28, 56372.
Venkataraman C., Lyons J. M., and Friedlander S. K. (1994) Size distributions of polycyclic
aromatic hydrocarbons and elemental carbon. 1. Sampling, measurement methods, and
source characterization. Environ. Sci. Technol. 28, 55562.
Verwer J. G. (1994) GaussSeidel iteration for stiff ODEs from chemical kinetics. SIAM J. Sci.
Comput. 15, 124350.
Villars D. S. (1959) A method of successive approximations for computing combustion equi-
libria on a high speed digital computer. J. Phys. Chem. 63, 5215.
Visser J. (1972) On Hamaker constants: A comparison between Hamaker constants and
Lifshitzvan der Waals constants. Adv. Colloid Interface Sci. 3, 33163.
Waggoner A. P., Weiss R. E., Ahlquist N. C., Covert D. S., Will S., and Charlson R. J. (1981)
Optical characteristics of atmospheric aerosols. Atmos. Environ. 15, 1891909.
Wagman D. D., Evans W. H., Parker V. B., et al. (1982) The NBS tables of chemical thermo-
dynamic properties: Selected values for inorganic and C
1
and C
2
organic substances in SI
units. J. Phys. Chem. Ref. Data 11, Suppl. 2.
Walcek C. (2000) Minor ux adjustment near mixing ratio extremes for simplied yet highly
accurate monotonic calculation of tracer advection. J. Geophys. Res. 105, 933548.
Walcek C. and Aleksic N. M. (1998) Asimple but accurate mass conservative, peak-preserving,
mixing ratio bounded advection algorithmwith Fortran code. Atmos. Environ. 32, 3863
80.
Walcek C. J., Brost R. A., and Chang J. S. (1986) SO
2
, sulfate and HNO
3
deposition velocities
computed using regional landuse and meteorological data. Atmos. Environ. 20, 949
64.
Walcek C. J., Yuan H.-H., and Stockwell W. R. (1997) The inuence of aqueous-phase chemical
reactions on ozone formation in polluted and nonpolluted clouds. Atmos. Environ. 31,
122137.
Walmsley J. L. and Wesely M. L. (1996) Modication of coded parameterizations of surface
resistances to gaseous dry deposition. Atmos. Environ. 30A, 11818.
Wang C. and Prinn R. G. (2000) On the roles of deep convective clouds in tropospheric
chemistry. J. Geophys. Res. 105, 22,26997.
Wang P. K., Grover S. N., and Pruppacher H. R. (1978) On the effect of electric charges on the
scavenging of aerosol particles by clouds and small raindrops. J. Atmos. Sci. 35, 173543.
Wanninkhof R. (1992) Relationship between wind speed and gas exchange over the ocean.
J. Geophys. Res. 97, 737382.
Washington W. M. and Parkinson C. L. (1986) An Introduction to Three-Dimensional Climate
Modeling. Mill Valley, CA, University Science Books.
Watson R. T. (1977) Rate constants for reactions of ClO
x
of atmospheric interest. J. Phys.
Chem. Ref. Data 6, 871917.
Weeks J. L. and Rabani J. (1966) The pulse radiolysis of deaerated aqueous carbonate solutions.
J. Phys. Chem. 70, 21006.
Weingartner E., Burtscher H., and Baltensperger U. (1997) Hygroscopic properties of carbon
and diesel soot particles. Atmos. Environ. 31, 231127.
Weisman M. L., Skamarock W. C., and Klemp J. B. (1997) The resolution dependence of
explicitly modeled convective systems. Mon. Wea. Rev. 125, 52748.
Welch R. M., Cox S. K., and Davis J. M. (1980) Solar Radiation and Clouds, Meteorological
Monograph 17. American Meteorological Society.
781
References
Wengle H. and Seinfeld J. H. (1978) Pseudospectral solution of atmospheric diffusion prob-
lems. J. Comp. Phys. 26, 87106.
Wentzel, M., Gorzawski, H., Naumann, K.-H., Saathoff, H., and Weinbruch, S. (2003) Trans-
mission electron microscopical and aerosol dynamical characterization of soot aerosols.
J. Aerosol Sci. 34, 134770.
Wesely M. L. (1989) Parameterization of surface resistances to gaseous dry deposition in
regional-scale numerical models. Atmos. Environ. 23, 1293304.
Wesely M. L. and Hicks B. B. (1977) Some factors that affect the deposition rates of sulfur
dioxide and similar gases on vegetation. J. Air Pollut. Control Ass. 27, 11106.
West R., Crisp D., and Chen L. (1990). Mapping transformation for broadband atmospheric
radiation calculations. J. Quant. Spectrosc. Radiat. Transfer 43, 1919.
Wetzel P. J. and Chang J. (1987) Concerning the relationship between evapotranspiration and
soil moisture. J. Climate Appl. Meteor. 26, 1827.
Wexler A. S. and Clegg S. L. (2002) Atmospheric aerosol models for systems including the
ions H
+
, NH
4
+
, Na
+
, SO
4
2
, NO
3
, Cl
, Br
, and H
2
O. J. Geophys. Res. 107 (D14)
10.1029/2001JD000451.
Wexler A. S. and Seinfeld J. H. (1990) The distribution of ammonium salts among a size and
composition dispersed aerosol. Atmos. Environ. 24A, 123146.
(1991) Second-generation inorganic aerosol model. Atmos. Environ. 25A, 273148.
Whitby E. R. (1985) The Model Aerosol Dynamics Model. Part I. Report to the U.S. Envi-
ronmental Protection Agency. Department of Mechanical Engineering, University of
Minnesota, Minneapolis.
Whitby K. T. (1978) The physical characteristics of sulfur aerosols. Atmos. Environ. 12, 135
59.
White M. (2000) Leonardo: The First Scientist. London, Abacus, 370pp.
Whitten G. Z., Hogo H., and Killus J. P. (1980) The carbon bond mechanism: A condensed
kinetic mechanism for photochemical smog. Environ. Sci. Technol. 14, 690700.
Wicker L. J. and Skamarock W. C. (1998) A time-splitting scheme for the elastic equations
incorporating second-order RungeKutta time differencing. Mon. Wea. Rev. 126, 1992
9.
(2002) Time-splitting methods for elastic models using forward time schemes. Mon. Wea.
Rev. 130, 208897.
Wilke C. R. and Chang P. (1955) Correlation of diffusion coefcients in dilute solutions.
Am. Inst. Chem. Eng. J. 1, 26470.
Wine P. H., Tang Y., Thorn, R. P., Wells J. R., and Davis D. D. (1989) Kinetics of aqueous-phase
reactions of the SO
4
radical with potential importance in cloud chemistry. J. Geophys.

Res. 94, 108594.
Wiscombe W. (1977) The delta-M method: Rapid yet accurate radiative ux calculations for
strongly asymmetric phase functions. J. Atmos. Sci. 34, 140822.
Wolf M. E. and Hidy G. M. (1997) Aerosols and climate: Anthropogenic emissions and trends
for 50 years. J. Geophys. Res. 102, 11, 11321.
Woodcock A. H. (1953) Salt nuclei in marine air as a function of altitude and wind force. J.
Meteorol. 10, 36271.
Woods T. N., Prinz D. K., Rottman, G. J., et al. (1996) Validation of the UARS solar ultraviolet
irradiances: Comparison with the ATLAS 1 and 2 measurements. J. Geophys. Res. 101,
954169.
World Meteorological Organization (WMO) (1975) Manual on the Observation of Clouds
and Other Meteors. World Meteorological Organization, Geneva.
(1995) Scientic Assessment of Ozone Depletion: 1994. Report 25, Global Ozone Research
and Monitoring Project, World Meteorological Organization, Geneva.
(1998) Scientic Assessment of Ozone Depletion: 1998. Report 44, WMO Global Ozone
Research and Monitoring Project, World Meteorological Organization, Geneva.
782
References
Worsnop D. R., Fox L. E., Zahniser M. S., and Wofsy S. C. (1993). Vapor pressures of
solid hydrates of nitric acid: Implications for polar stratospheric clouds. Science 259,
714.
Wu J. (1993) Production of spume drops by the wind tearing of wave crests: The search for
quantication. J. Geophys. Res. 98, 18,2217.
Wu Y.-C. and Hamer W. J. (1980) Revised values of the osmotic coefcients and mean activity
coefcients of sodium nitrate in water at 25

C. J. Phys. Chem. Ref. Data 9, 5138.
Xiong C. and Friedlander S. K. (2001) Morphological properties of atmospheric aerosol aggre-
gates. Proc. Natl. Acad. Sci. 9, 11,8516.
Yabe T., Tanaka R., Nakamura T., and Xiao F. (2001) An exactly conservative semi-Lagrangian
scheme (CIP-CSL) in one dimension. Mon. Wea. Rev. 129, 33244.
Yamamoto G., Tanaka M., and Asano S. (1970) Radiative transfer in water clouds in the
infrared region. J. Atmos. Sci. 27, 28292.
Yamartino R. J. (1993) Nonnegative conserved scalar transport using grid-cell-centered spec-
trally constrained Blackman cubics for applications on a variable-thickness mesh. Mon.
Wea. Rev. 121, 75363.
Yanenko N. A. (1971) The Method of Fractional Steps. Berlin, Springer-Verlag, 160pp.
Yin F., Grosjean D., and Seinfeld J. H. (1990) Photooxidation of dimethyl sulde and dimethyl
disulde. I: Mechanism development. J. Atmos. Chem. 11, 30964.
Young A. T. (1980) Revised depolarization corrections for atmospheric extinction. Appl. Opt.
19, 34278.
Young T. R. and Boris J. P. (1977) A numerical technique for solving stiff ordinary differential
equations associated with the chemical kinetics of reactive-ow problems. J. Phys. Chem.
81, 24247.
Zawadski I., Torlaschi E., and Sauvageau, R. (1981) The relationship between mesoscale
thermodynamic variables and convective precipitation. J. Atmos. Sci. 38, 153540.
Zaytsev I. D. and Aseyev G. G., eds. (1992) Properties of Aqueous Solutions of Electrolytes
(translated by M. A. Lazarev and V. R. Sorochenko). Boca Raton, FL, CRC Press.
Zeldovich Y. B. (1942) Theory of new-phase formation: cavitation. J. Exp. Theor. Phys. (USSR)
12, 52538.
Zhang D. and Anthes R. A. (1982) A high-resolution model of the planetary boundary layer
sensitivity tests and comparisons with SESAME-79 data. J. Appl. Meteor. 21, 1594609.
Zhang Y., Bischof C. H., Easter R. C., and Wu P.-T. (1998) Sensitivity analysis of multi-phase
chemical mechanism using automatic differentiation. J. Geophys. Res. 103, 18, 95379.
Zhang Y., Seigneur C., Seinfeld J. H., Jacobson M. Z., and Binkowski F. (1999) Simulation of
aerosol dynamics: Acomparative reviewof algorithms used in air quality models. Aerosol
Sci. Technol. 31, 487514.
Zhang Y., Seigneur C., Seinfeld J. H., Jacobson M., Clegg S.L, and Binkowski F. (2000) Acom-
parative review of inorganic aerosol thermodynamic equilibrium modules: Similarities,
differences, and their likely causes. Atmos. Environ. 34, 11737.
Zhang Y., Pun. B., Wu S.-Y., et al. (2004) Development and application of the model for
aerosol dynamics, reaction, ionization and dissolution (MADRID). J. Geophys. Res. 109,
D01202, doi: 10.1029/2003JD 003501.
Zhao J. and Turco R. P. (1995) Nucleation simulations in the wake of a jet aircraft in strato-
spheric ight. J. Aerosol Sci. 26, 77995.
783
Index
A
absolute temperature, see temperature, absolute
absorption, gas by liquid, 408
absorption, radiative, 276, 283, 287, 290305,
312, 539, 692
gas, 2904, 300, 313, 340
Mie-regime, 308
particle, 290313
Tyndall-regime, 306
absorption coefcient, mass, 295
absorption cross section, see cross section,
absorption
absorption efciency, particle, 305
absorptivity, 277
abstraction, chemical, 370
acceleration
centripetal, Earth, 82, 89, 95
centripetal, small scale, 89
Coriolis, 82, 89, 90, 935, 97, 98, 108, 110,
113, 127, 128, 135, 489
gravitational, 13, 34, 82, 90, 959, 109, 135,
475, 519, 661, 662
effective, 95, 98
globally averaged, 98, 715
true, 95
vertical prole, 71415
inertial (total), 87, 88
local, 82, 89, 903
accommodation coefcient
mass, 411, 530, 531
thermal, 532
accumulation mode, 308, 458, 459, 460, 480
composition, 482
accuracy of numerical solution, 420
accurate space derivative (ASD) method, 200
acetaldehyde, gas, 365
production of ozone, 3669
structure, 392
volume mixing ratio, 444
acetic acid, aqueous, 448, 739
acetic acid, gas, 444
acetone, 366, 369
e-folding lifetime, 369, 382
photolysis coefcient, 352
volume mixing ratio, 369, 444
acetylene, see ethyne
acid, 338, 556
strong, 556
weak, 556, 557
acid deposition, 6, 7, 646, 652, 681, 682
acid fog, 7, 645
acidication, sea spray, 470, 566
acidity, 555, 646, 647, 659
acoustic cutoff frequency, 125, 126, 128
acoustic waves, see waves
actinic ux, 10, 11, 279, 320, 333, 351, 693
activation energy, see energy, activation
active chlorine, 404, 405, 409, 413, 414
activity, thermodynamic, 5589
activity coefcient, solute, 559
binary, 563, 5657
data, 568, 572, 577, 693, 7468, 749
temperature dependence, 5678, 572
geometric mean, 563
mixed, 563, 56870
single-ion, 563
AdamsBashforth scheme, see time-stepping
schemes, AdamsBashforth
addition, chemical, 370
adiabatic lapse rate, see lapse rate
adiabatic process, 47, 49, 50, 52, 53, 59, 122,
131, 132, 134, 139, 602, 603, 6057
compression, 50
expansion, 50
moist, 602
pseudoadiabatic, 6027, 613, 643
adsorption, 408
characteristic time of, 491
heat of, see enthalpy, of adsorption
rate, 411
advection, 22, 69, 171, 192, 199
advectiondiffusion equation, 169, 1712,
181202
advection equation, 181
advection fog, see fog, advection
advection schemes, 199202
aerobic, 15
aerodynamic radius, see radius, aerodynamic
aerodynamic resistance, see resistance,
aerodynamic
784
Index
aerodynamic roughness, see surface roughness
length, for momentum
aerosol, 8, 446
aerosol particle, see particle, aerosol
ageostrophic, 132
air density, see density, moist air
air pollutant, dened, 7
air pollution history of, 56
air pressure, see pressure, air
air quality data, see ambient data
air quality model, see model, air quality
airsea exchange of gases, 6729
numerical solution, 6737
simulations of, 6789
albedo, 11, 22, 283, 284, 329, 330, 331, 334,
693, 695
clay, 22
cloud, 284
desert, 284
forest, 284
grass, 284
planetary (Earth-atmosphere), 283, 326
sand, 22
snow, 284
UV-A, 284
UV-B, 284
water, 284
Alberti, Leone Battista, 1
alcohol, 338
ozone production from, 3901
aldehyde, 338
Alkali Act of 1863, 6
alkaline, 555
alkane, 338
alkene, 338
alkyne, 338
alternating-directions scheme, 172
altimeter, radar, 36
altitude coordinate, see coordinates, altitude
altocumulus, see cloud types, altocumulus
altostratus, see cloud types, altostratus
aluminum oxide, 302, 448, 480, 740
ambient data, 6934
ammonia, aqueous, 448, 557, 587, 594, 651, 739,
751
ammonia, gas, 555, 558
ammonication, 15
ammonium bisulfate, 448, 739
ammonium chloride, 448, 558, 594, 596, 739
ammonium ion, 15, 448, 482, 534, 535, 557, 589,
590, 594, 651, 739
ammonium nitrate, 302, 448, 558, 574, 582, 583,
593, 594, 739
ammonium sulfate, 302, 303, 334, 335, 448, 449,
534, 535, 538, 570, 575, 582, 583, 739
anaerobic, 15, 17, 373
analytical predictor of condensation (APC)
scheme, 5445
analytical predictor of dissolution (APD) scheme,
585, 586
analytical predictor of dissolution and dissociation
(APDD), 585
analytical solution, 421
anelastic approximation, 138, 139, 140, 142, 160,
163, 166
anelastic continuity equation, 142
anemometer
pressure-plate, 1
pressure-tube, 2
spinning-cup, 1, 2
swinging-plate, 1
angular velocity, 200
Earth, see Earth, angular velocity
small-scale wind, see wind scalar velocity,
angular
anion, 554, 556, 565, 569, 597
anisotropic correction term, 301
Antarctic ozone hole, see ozone, hole, Antarctic
anticyclone, 115
antidiffusion velocity, 200
apparent force, see forces
apparent molal enthalpy, see enthalpy, apparent
molal
apparent molal heat capacity, see heat capacity,
molal, apparent
aqueous chemistry, 64559, see also chemical
reactions, aqueous
aragonite, 677, see also calcium carbonate
Arakawa C-grid, 209
ArakawaSchubert scheme, 613, 614
area-equivalent radius, see radius, area-equivalent
argon, 12, 14, 15
Aristotle, 34
aromatic, 338
aromatic polycarboxylic acids, 302
arrays, computer
global, 688
work, 688
Arrhenius, Svante, 347
Arrhenius equation, 347, 561
Ash, 478, 482, see also y ash
asphalt, 265
emissivity, see emissivity, asphalt
temperature, see temperature, road
thermal conductivity, see conductivity, thermal,
asphalt
asphalt shingles, 266
asymmetry parameter
effective, 329
single-particle, 323
atmospheric window, 294
Avogadro, Amedeo, 29
Avogadros law, 28
785
Index
Avogadros number, 29
azimuth angle, 278, 318
B
backscatter ratio, 324
backscattering, 287
backsubstitution, see matrix
backward difference, see nite-difference method,
backward difference
backward differentiation formula, 432
backward Euler, see nite-difference method,
backward Euler
bacteria, 481
barometer
aneroid, 2, 35
mercury, 2
base, 557, 585, 586, 613, 614
basic, 555
basis function, 193, 194, 199, 200
Beers law, 327
benzaldehyde, 388
structure, 392
benzaldehydes, 302
benzene, 392
benzoic acids, 302
benzylperoxy radical, 387
Bergeron process, 43
beta plane approximation, 132
bicarbonate ion, 448, 739
binary activity coefcient, see activity coefcient,
solute, binary
biogenic emissions, 388
biofuel burning, 7, 341, 376, 380, 479, 480
biomass burning, 17, 339, 386, 396, 470, 474,
478, 479, 693
bisulfate ion, 448, 589, 646, 739
bisulte ion, 448, 646, 648, 657, 739
bisulte radical, gas, 375
bitumen, 267
Bjerknes, Vilhelm, 4
black carbon, 276, 290, 294, 302, 307, 311, 312,
446, 448, 449, 458, 459, 466, 477, 478,
479, 480, 482, 483, 536, 693, 701, 702,
see also soot
blackbody, 276, 277, 278, 279, 280, 282, 283,
305, 326
Boltzmanns constant, 18, 29
Bond number, 664
boundary conditions, see model, boundary
conditions
boundary layer, 1924, 22870
cloud layer, 24
entrainment zone, 23, 24, 114
mixed layer, convective, 23
nocturnal (stable), 24
residual layer, 24
subcloud layer, 24
surface layer, 23
boundary value problem, 170
bounded numerical solution, 199, 200, 209, 545,
585
Boussinesq approximation, 138, 1423
box model, see model, zero-dimensional
Boyle, Robert, 28
Boyles law, 28
breakup, drop, 6223
broadening factor, 349
bromine monoxide, 406, 407, 444
bromine nitrate, 406
e-folding lifetime, 407
Bromleys method, 569
bromocarbons, 400, 401
brown cloud, 299
Brownian coagulation, 508
collision kernel, see collision kernel,
Brownian
Brownian diffusion, 507, 50810, 511, 516, 519,
521, 522
kernel, see coagulation kernel
Brownian motion, 508, 511, 516, 519, 521, 522,
523, 622
Bruggeman mixing rule, 311
BruntV ais al a frequency, 567
buffer zone, 691
bulk aerodynamic formula, 2379
bulk parameters, 77
bulk Richardson number, see Richardson number,
bulk
bulk transfer coefcient
energy, see heat-transfer coefcient
water vapor, 239
buoyancy, 22, 23, 24, 57, 73, 89, 105, 106, 118,
127, 128, 129, 142, 228, 229, 235, 242,
243, 244, 245, 251, 252, 254, 600, 601,
607, 611, 642, 662, see also turbulence,
thermal
production rate of, 253
buoyancy factor, 610, 611
buoyancy oscillation, see oscillation, buoyancy
buoyancy wave, see waves, gravity
butane, 369, 379, 381, 382, 391, 392, 720
structure, 392
butene, 382
structure, 392
C
C-grid, Arakawa, 209
calcite, 677, see also calcium carbonate
calcium carbonate, 302, 448, 474, 740, 741
calcium chloride, 448, 740
calcium ion, 448, 479, 482, 554, 739
calcium nitrate, 448, 739
calcium sulfate, 448, 740
786
Index
canopy, 233, 234, 249, 250, 259, 260, 262, 264,
265, 271, 668, 669, 670, 679
canopy height, 233, 249
capacitance, electrical, 549, 550, 615
carbohydrates, 15, 16
carbon bond lumping mechanism, 391
carbon dioxide, gas, 7, 12, 14, 15, 1617, 25,
191, 2914, 3415, 349, 350, 355, 3608,
383, 384, 449, 477, 529, 557, 588, 668,
673, 676, 679, 682, 738, 739
airsea ux, 6729
structure, 341
volume mixing ratio, 17, 292, 341, 444
carbon disulde, 370
carbon monosulde, 374
carbon monoxide, gas, 14, 15, 16, 17, 30, 2913,
297, 311, 341, 355, 361, 363, 367, 375,
379, 380, 384, 398, 444, 448, 474, 482,
529, 5547, 588, 593, 672, 67580,
73841, 7446
structure, 341
carbon tetrachloride, 292
carbonate ion, 448, 482, 483, 739
carbonic acid, 448, 449, 557
carbonyl, gas, 338
carbonyl sulde, 370
Cartesian coordinates, see coordinates, Cartesian
catalytic ozone destruction, see ozone, catalytic
destruction cycles
cation, 554, 565, 569, 677
cavity ringdown method, 347
Celsius, Anders, 2
central difference approximation, see
nite-difference method, central difference
CFC-11, 291, 292
CFC-12, 291, 292
CFC-113
chain length, 396, 397, 404, 406
chapeau function, 194
Chapman, Sydney, 27, 528
Chapman cycle, 27
Chappuis, M. J., 291
Chappuis bands, 291
charge, particle, 640
charge balance, see conservation, charge
charge density, 639
charge separation, 639, 640, 641
Charles law, 28, 29
Charles, Jacques, 29
Charney, Jule, 5
CharneyPhillips grid, 205
Charnock constant, 232
Charnock relation, 232
chemical equation solvers, 41842
chemical equilibrium, see equilibrium
chemical loss rate, see chemical reactions, gas, loss
chemical mechanisms
condensed, 3913
carbon-bond lumping, 391
lumped species, 391, 393
surrogate species, 391, 393
explicit
chemical potential, 560
standard, 560
chemical production rate, see chemical reactions,
gas, production
chemical rate coefcient, 344, 34651
data, 692, 72838
rst-order, 344
heterogeneous reactions, 346, 348, 353, 409,
746
high-pressure limit, 349
low-pressure limit, 349
photolysis reactions, see photolysis
pressure dependence, 34950
pseudo-rst-order, 344
pseudo-second-order, 344
second-order, 344
temperature dependence, 3478
third-order, 344
chemical reactions, aqueous, 10, 76, 645, 654
list of, 7436
rate of, 654, 656
chemical reactions, gas, 3424
bimolecular, 342
combination, 343
elementary, 342
equilibrium, see equilibrium, reactions
heterogeneous, 10, 74, 362
homogeneous, 362
isomerization, 343
kinetic (collision), 9, 342
list of, 72838
nighttime, 3623, 400
nonelementary, 343
photochemical, 9, 336, 357
photolysis, see photolysis
pressure-dependent, 343
rate, 3446
loss, 344, 345, 353, 354, 428, 656, 657
production, 63, 353, 428, 484, 642, 655
reversible, 362, 554, 560
787
Index
chemical reactions, gas (cont.)
sets of, 3513
stratospheric, 393415
termolecular, 342
thermal decomposition, 343
tropospheric, 35775
unimolecular, 342
urban, 37593
chemical structures, see structures, chemical
chloride ion, 448, 479, 482, 534, 554, 566, 573,
575, 586, 589, 594, 596
chlorine, active, 404, 405, 409, 413, 414
chlorine, molecular, 291, 399, 400, 409, 412, 416,
444, 477
chlorine activation, 409, 413
chlorine monoxide, 355, 404, 405, 409, 410, 413,
444
chlorine nitrate, 405, 416
chlorine reservoirs, 405, 413
chlorite, 474
chlorocarbons, 400, 401
chlorodiuoromethane, see HCFC-22
chlorouorocarbons (CFCs), 291, 294, 399, 400,
401, 402, 403, 404, 405, 414
chlorophylls, 15
chloroplasts, 668
cirrocumulus, see cloud types, cirrocumulus
cirrostratus, see cloud types, cirrostratus
cirrus, see cloud types, cirrus
classical nucleation theory, 48491, 493
ClausiusClapeyron equation, 12, 403, 400, 526,
603
clay, 21, 257, 267, 474
clay loam, 257
climate, dened, 7
climate change, see global climate change
closure, order of, see order of closure
closure scheme
local, 252
nonlocal, 250, 252
cloud condensation nuclei (CCN), 536, 537, 538,
548, 549, 551, 615, 616, 619, 634, 635
activation, 536, 616
cloud drop, 9, 10, 43, 273, 289, 290, 308, 312,
321, 333, 369, 374, 446, 447, 449, 494,
511, 518, 519, 520, 536, 538, 581, 595,
614, 615, 6459, 6526, 659, 661, 663
diameter, 447
number concentration, 447
mass concentration, 447
cloud formation, 53, 595, 6001, 6067
cloud fraction, 640
cloud heights, 598, 599
cloud layer, see boundary layer, cloud layer
cloud microphysics, 598642
cloud thermodynamics, 598642
cloud types, 598601
altocumulus, 601
altostratus, 601
cirrocumulus, 601
cirrostratus, 601
cirrus, 277, 599, 601
cold, 43
convective, 150, 600
cumulonimbus, 601, 612
cumulus, 8, 511, 599, 601, 607, 608, 612,
613
nimbostratus, 461, 600
nimbus, 599, 606
orographic, 601
polar stratospheric, 40813
type I, 40810
type II, 40810
stratocumulus, 601
stratus, 599
warm, 620
coagulation, 10, 20, 28, 39, 43, 76, 77, 235, 446,
447, 45868, 481, 482, 485, 487, 490,
493, 494524, 5306, 54750, 553, 598,
610, 614, 61925, 63543, 652, 664, 665,
689
aerosolaerosol, 10, 465, 466, 4968, 5005
aerosolhydrometeor, 10, 519, 6346
effects of condensation on, 465, 467, 5478
fully implicit equation, 4947
hydrometeorhydrometeor, 10, 6202
implicit solution, 4947, 508
importance of, 494
integro-differential equation, 494
multiple size distribution interactions, 5005
semiimplicit solution, 465, 466, 4968, 501
self-preserving solution, 500
Smoluchowskis solution, 499, 523, 524
coagulation kernel, 50822, 622, see also collision
kernel
coal, 479, 480
anthracite, 479
bituminous, 479
lignite, 479
peat, 479
coalescence efciency, 508, 5212, 523, 622, 639,
641, 643
coarse-particle mode, 458, 460
composition, 482
coefcient of drag, see drag
cold front, 601
collection efciency, see coalescence efciency
collision-coalescence, see coagulation,
hydrometeorhydrometeor
collision complex, 343
collision diameter, gas, see gas, diameter
collision efciency, 510, 521
collision kernel, 508, 509, 510, 511, 512, 515,
516, 517, 522, 524, 622, 639, 641, 643
788
Index
Brownian, 508
diffusion enhancement, 510
diffusiophoresis, 51721
electric charge, 101, 273, 448, 508, 51721,
522, 524, 622
fractal correction, 51416
gravitational collection, 51011
thermophoresis, 51721
turbulent inertial motion, 294, 51112, 516
turbulent shear, 294, 508, 51112, 516, 519,
521, 522, 524, 622
van der Waals correction, 513
collision radius of particle, see radius, collision
collisional prefactor, 347
cooling rate, drop, 526
combination reaction, see chemical reactions, gas,
combination
combustion reaction, 380
compressible uid, 67, 122, 135, 138, 139,
140
computer
memory, 685
optimization, 432, 436, 6889
array minimization, 437
grid blocking, 438
sparse-matrix techniques, 432, 434, 436,
543, 659, 688
vectorization, 436
condensation, 10, 436, 74, 80, 447, 448, 482,
52550, 553, 556, 570, 607, 636, 642,
645, 646, 647, 649, 653, 669
effects on coagulation, 465, 467, 5478
effects on homogeneous nucleation,
5457
growth equation, 525, 5401
mass ux form, 527
multiple component form, 540
radius ux form, 528
volume concentration form, 541
volume ux form, 528
simulations of, 545
solution to, 5423, 545: with homogeneous
nucleation, 5457
hyrometeor formation from, 61420
conditional instability, see stability, atmospheric
conduction, 201, 255
conductive heat ux, 21, 256, 258
conductivity
hydraulic, 257
thermal, 20, 235, 255, 256, 257, 259, 270, 271,
272, 520, 526, 527, 528, 531, 532, 533,
550, 615, 629, 633, 634
asphalt, 266
clay, 20
dry air, 20
dry sand, 20
ice, 269
liquid water, 20, 43, 257, 277, 284, 668
moist air, 531, 532: corrected, 531, 532; for
collision geometry, 532; for sticking
probability, 530
particle, 520
soilwaterair mixture, 256
snow, 268, 713
water vapor, 20
consistency, 159, 173, 174, 175, 208, 215, 217,
311, 415, 679
conservation
charge, 564, 576, 577, 588, 589, 590,
596
mass, 420
mole, 541, 576, 587, 637, 654
constants, 71113
constructive interference, 287
contact angle, 489, 490
contact freezing, 6245
continuity equation, 6578
advective form, 66
anelastic, 142
ux divergence form, 66
for air, 4, 11, 61, 67, 72, 74, 78, 79, 105,
106, 111, 122, 124, 132, 135, 13847,
1547, 162, 167, 168, 204, 208, 214, 215,
224
altitude coordinate, 66, 71, 74
elastic, 165
pressure coordinate, 147
sigma-altitude coordinate, 1623
sigma-pressure coordinate, 1547
for energy, 79
for gases/trace species, 11, 66, 69, 71, 745,
836, 111, 124, 140, 143, 147, 149, 152,
157, 16772, 196, 204, 209, 21113, 215,
216, 219, 224, 226, 312, 416, 493, 638,
685, 688, 6947, 702
altitude coordinate, 66
moist-air mass mixing ratio, 67, 72
for incompressible uid, 67
for particles, 75
volume concentration, 76
for water
ice, 77
liquid, 77
vapor, 77
total water, 11
velocity divergence (advective) form, 66
continuum ow, 663
continuum regime, 5068, 513, 516, 525, 530
contour lines, 150, 151
contrast ratio, 314, 315
liminal, 314
789
Index
convection, 21
forced, 21, 22, 23, 34, 54, 59, 101, 102, 105,
124, 126, 131, 199, 205, 251, 341, 355,
601, 607
free, 4, 10, 15, 22, 53, 233, 244, 251, 273, 275,
276, 282, 28592, 300, 3058, 31115,
334, 340, 357, 375, 399, 407, 421, 446,
458, 478, 509, 600, 6067, 639, 642, 666,
702, 711
convective available potential energy (CAPE), 612
convective clouds, see cloud types, convective
convective plume scheme, 251
convergence, 22, 83, 173, 174, 427, 430, 431,
577, 578, 579, 601
coordinates, 62, 824, 107, 110, 115
altitude, 62, 143
Cartesian (rectangular), 61, 82, 116
cylindrical, 115, 116, 117, 136
isentropic, 138
Lambert conformal, 83
Mercator, 83
pressure (isobaric), 143
sigma-altitude, 85, 138, 1606
sigma-pressure, 85, 138, 15160
spherical, 82, 83
stereographic, 83
universal transverse Mercator (UTM), 82
vertical, 85, 13866, 167, 204, 225
Coriolis force, see acceleration, Coriolis
Coriolis parameter, 95, 128, 131, 132, 227
corona, 287
correlated k-distribution method, 298
multiple-absorber, 298
Coulomb force, see forces, electrostatic
Coulombs constant, 638, 641
Coulombs law, 638
CourantFriedrichsLewy (CFL) criterion, 1812
covalent bond, 336
coverage, surface, see surface coverage
CrankNicolson parameter, 184
CrankNicolson scheme, 169, 1845, 191, 195,
200, 222
creep, 475
cresol, 388
criegee biradical, 384
critical angle, see reection, critical angle
critical radius
for growth, 536, 537, 538, 616, 617
for nucleation, 48591
critical Richardson number, 245
critical saturation ratio, 536, 537, 538, 616, 617
cross section
absorption
gas, 294, 333, 692
particle, 305
scattering
gas, 300
particle, 305
Cryfts, Nicolas, see Cusa, Nicolas of
crystallization, 554, 575, 591, 594, 677
crystallization relative humidity (CRH), 575, 593
cubic spline method, 192, 542
cumulonimbus, see cloud types, cumulonimbus
cumulus, see cloud types, cumulus
cumulus parameterizations, 598, 61214
ArakawaSchubert, 613, 614
Kuo, 613
moist convective adjustment, 613
Cunningham slip-ow correction, 508, 509, 519,
662, 663
curvature, Earths, see Earth, curvature
curvature effect, see Kelvin effect
Cusa, Nicolas of, 2
cuticle, plant, 668, 669, 679
cutin, 668
cutoff level, statistical, 697
cycloalkane, 338
cyclohexene, 392
cyclone, 114
cyclopentane, 392
cyclopentene, 392
cylindrical coordinates, see coordinates,
cylindrical
D
Dalton, John, 30
Daltons law of partial pressure, 30
data assimilation, 690
da Vinci, Leonardo, 2
debugging, 699
DebyeH uckel parameter, 569
declination angle, solar, 317, 319
decomposition, biological, 16
decomposition, matrix, see matrix, solution
methods, decomposition
degree of equation, 169, 170, 171
dehydration, 471
deliquescence, 574
deliquescence relative humidity (DRH), 5745
mutual, 575, 582
delta functions, 332
dendrite, 44
denitrication, microbial, 15
denitrication, stratospheric, 409
density
dry air, 30
ice, 270
liquid water, 67, 460
moist air, 11, 12, 13, 27, 33, 34, 36, 66, 67, 70,
71, 78, 81, 101, 104, 106, 108, 140, 147,
163, 206, 207, 229, 230, 232, 260, 475,
493, 506, 520, 523, 524, 528, 530, 532,
551, 663, 666, 714
particle, 453
soilwaterair mixture, 256
seawater, 269
790
Index
solid soil, 256
snow, 268
water vapor, 31
at saturation, 41
dependent variable, 143, 144, 169, 170
deposition, dry, 661, 66572
gas, 10
speed, 665
particle (or aerosol), 10
speed, 666
deposition, ice or solid, 10, 16, 436, 80,
52550
growth equations for, 54850
hydrometeor formation from, 61420
derivative
local, 63, 78, 138, 141
total, 63, 65, 77
Descartes, Rene, 3
desorption, 409, 411, 533
rate of, 411
destructive interference, 287
des Voeux, Harold Antoine, 5
detrainment, 608
dew, 599, 669
frozen, 599
dew point, 45, 46, 59, 602, 606, 607, 701
diabatic energy sources/sinks, 47, 50, 59, 80, 81,
122, 226, 332, 609
diabatic heating rate, see heating rate, diabatic
diagnostic equation, 140, 142, 143, 166, 208,
210, 221, 244, 253, 254, 683
for nonhydrostatic pressure, 142
for vertical velocity, 157
diamond dust, 599
dichlorine dioxide, 412
dichlorodiuoromethane, see CFC-
dielectric, 309, 310, 311, 334
dielectric constant, 309
complex, 310
diesel fuel, 479
diffraction, 2867, 408
diffuse radiation, 320, 321, 330, 334
diffusion
eddy, 105, see also turbulence
gas in air, 8, 68, 71, 102, 107, 409, 528, 531,
668
molecule in water, 6524
time scale, 652
numerical, see numerical diffusion
particle in air, 508
diffusion coefcient
aqueous, 653
eddy, 73, 237, 666, 667
for energy (eddy thermal diffusivity), 73, 79,
239, 247, 251, 271: tensor, 74
for momentum (eddy viscosity), 73, 106, 245:
tensor, 107
for trace gases, 247
for water vapor, 247
molecular
for air, see viscosity, kinematic
for energy (molecular thermal diffusivity),
233
for trace gases, 68, 102, 525, 527, 528, 529,
531, 550: corrected, 529; for collision
geometry, 530, 550; for sticking
probability, 530: effective, 541
for water vapor, 233, 520: effective, 541
particle, 508
water in soil, 257
diffusion equation, 181
diffusiophoresis, 517
diffusiophoresis
dimer mechanism, 412
dimethyl disulde (DMDS), 369
dimethyl sulde (DMS), 336, 369
volume mixing ratio, 14, 16, 31, 32, 58, 334,
346, 415, 444
dimethyl sulfone (DMSO), 372
Dines, William Henry, 2
dinitrogen pentoxide, 291
dipoledipole interactions, 409, 512
direct numerical simulation, see model, direct
numerical simulation
direct radiation, see radiation, direct
directional derivative, see gradient operator
discrete size distribution, see size distribution,
discrete
dispersion, light, 1225, 127, 129, 132, 134, 135,
197, 283, 286
dispersion relationship, see waves, dispersion
relationship
dispersive refraction, see dispersion, light
displacement height, 232, 249, 250, 271
dissipation, 101, 235, 236, 252, 253, 254, 521,
523
of turbulent kinetic energy, see turbulent kinetic
energy, dissipation rate
dissociation, 448, 534, 535, 551, 554, 555, 556,
557, 563, 584, 617, 643, 645, 654, 655
reactions, 5557
dissolution, 10, 75, 76, 525, 55394, 595, 649,
676
reactions, 555
growth solutions
dissociatiating species, 58591
nondissociating species, 5835
distributed memory machine, 686, 687
diterpenes, 388
divergence, scalar, 64, 66, 67, 71, 74, 136, 141,
332
DMS, see dimethyl sulde
Dobson unit, 394
dolomite, 474
791
Index
domain of inuence, 694
Donora, Pennsylvania, 375
dot product, 64, 86
drag
form, 237
skin, 237
wave, 237
drag coefcient, 237, 238, 245
drag force, see forces, drag
drift speed, 519
drizzle, 447, 599
supercooled, 1, 599, 669
drop breakup, see breakup, drop
drop surface temperature, see temperature,
drop surface
dry air, dened, 30
dry deposition, see deposition, dry
dynamic viscosity of air, see viscosity, molecular,
air, dynamic
dynamical meteorology, 4, 6
E
Earth
angular velocity, 88
vector, 88, 97
cross-sectional area, 326
curvature, 82
effective temperature of, 281
equilibrium temperature, 326
mass of, 97, 712
radius of, 14, 85, 712
radius vector for, 88
shape, 84, 95
EarthSun distance, 325
ecliptic, 317, 318
ecliptic longitude of the Sun, 317
Eddington approximations
diffuse phase function, 330
solar phase function, 330
eddy, 8, 68, 73, 105, 235
large, 235, 236, 251
eddy diffusion, see diffusion, eddy
eddy thermal diffusivity, see diffusion coefcient,
eddy, for energy
eddy viscosity, see diffusion coefcient, eddy, for
momentum
Edinburgh, 375
effective medium approximations, see mixing
rule, optical
Einstein, Albert, 275
Einstein relation, 519
El Chich on, 398
Ekman number, 109, 110
electric charge, 520
collision kernel, see collision kernel, electric
charge
electric eld strength, 638, 639, 641, 643
threshold, 638
electrolytes, 555, 556, 557, 558, 56375, 616, 748
multivalent, 564
nonsymmetric, 564
solid, 557
strong, 556
symmetric, 64, 179, 226, 322, 564
univalent, 564
weak, 556
electromagnetic energy, see radiation
electron dot symbols, see Lewis symbols
electron pair, 337
electrostatic force, see forces, electrostatic
elemental carbon, see black carbon
elementary reaction, see chemical reactions gas,
elementary
elements in gases, 336
emission, 74
aerosol, see emission, particle
area source, 380
bacteria, 481
biofuel burning, see biofuel burning
biogenic, 388
biomass burning, see biomass burning
combustion, 5, 6, 16, 17, 37, 302, 336, 33941,
366, 375, 376, 380, 386, 396, 457, 470,
479, 480, 482, 650, 693
data, 693
dened, 9
fossil-fuel, 47980
forest re, 478, 479
fugitive, 477
industrial, 4801
inventories
gas, 379
particle, 483
meteoric debris, 481
mobile source, 380
particle, 47082
point source, 380
pollen, 481
rates, 379
resuspension, 477, 665
sea spray, 37, 4703
soil dust, 4737
ux, 475, 476
spores, 481
spume drop, 471
stationary source, 121, 380
tire particle, 481
viruses, 481
volcanic, 37, 477, 478
emissivity, 22, 260, 277, 302, 326, 331, 334
asphalt, 266
concrete, 277
desert, 234, 277, 281, 284, 302, 474, 750
792
Index
forest, 277
globally averaged, 326
grass, 277
ice, 269
liquid water, 269
soil, 277
snow, 268
vegetation, 262
empirical equation, 7, 240, 532
endothermic process, 47
energy, 37, 78, 238, 246, 273, 274, 709
activation
of adsorption, 411
of chemical reaction, 347
of desorption, 411
electromagnetic, see radiation
Gibbs free
of a chemical system, 559, 560, 576
of adsorption, 411
of desorption, 411
of heterogeneous nucleation, 490
of homogeneous nucleation, 484
standard molal, of formation, 561, 7401
gravitational potential, 37, 98
internal, 37, 47
kinetic, 18, 27, 37, 40, 49, 57, 68, 101,
104, 214, 215, 235, 342, 506, 521,
522, 528
turbulent, see turbulent kinetic energy
potential, 37
surface, 484
work, 37, 484
energy-balance equation, surface, 259
energy-containing subrange, 235
energy density, 79
energy transfer, 47
ENIAC, 5
enstrophy, 214
enthalpy, 559, 560, 567, 613
apparent molal, 567
at innite dilution, 568
relative, 7, 21, 24, 31, 32, 44, 45, 46
of adsorption, 491
of dilution, 567
of formation
molal, 561
standard molal, 561
entrainment, 60810
rate, 609, 613
zone, see boundary layer, entrainment zone
entropy, 52, 411, 559, 560
of adsorption, 411
envelope, see waves, envelope
environmental lapse rate, see lapse rate,
environmental
epidermis, leaf, 668
epsomite, 474
equations of atmospheric dynamics, 111
nite-difference solution to, 20424
equation of motion, see momentum equation
equation of state, 7, 11, 12, 28, 29, 31, 33, 36, 48,
57, 60, 111, 140, 143, 226, 526, 529
for dry air, 30
for moist air, 33
for water vapor, 31
Equator, 25, 26, 59, 83, 274, 275, 317, 318, 319,
550
equilibrium, 10, 55394
assumption, 583
coefcient, 55862
list of, 7413
temperature dependence, 5612
model
equations for, 5757
simulation results, 5823
solution methods: AEI, 57981, 597, 655;
MFI, 5779
reactions, 55465, 649, 7413
dissolution, see dissolution, reactions
dissociation, see dissociation, reactions
heterogeneous, 558
list of, 7413
precipitation, see precipitation, solid,
reactions
equilibrium saturation ratio, see saturation ratio,
equilibrium
equilibrium vapor pressure, see vapor pressure,
saturation
equinox, 319
equivalent depth, 129, 130, 135
equivalent potential temperature, see potential
temperature, equivalent
error, see normalized gross error
error test
global, 433
local, 433
error tolerance
absolute, 433
relative, 433
etage, see cloud heights
ethane, 365
structure, 392
ethanol, 390
ethanoloxy radical, see hydroxyethyloxy radical
ethene, 365
structure, 392
ethene molozonide, 384
ethoxy radical, 365
ethylbenzene, 392
793
Index
ethyl nitrate, 365
ethyl nitrite, 365
ethylperoxy nitric acid, 366
ethylperoxy radical, 365
ethyne (acetylene)
structure, 392
Eulerian, 63, 64, 111, 201, 462, 463, 466, 468,
494
evaporation, 10, 44, 74, 80, 274, 447, 448, 517,
518, 52550, 553, 556, 570, 613, 628,
631, 636, 645, 646, 649, 653, 669
from soil, 258
evaporation fog, see fog, evaporation
evaporative freezing, see freezing, evaporative
excited atomic oxygen, see oxygen, excited atomic
Exner function, 52
exothermic process, 47, 78
explicit equation/solution, 139, 140, 181, 182,
186, 187, 198, 211, 222, 41823, 442,
552, 584, 620
exponential method, 418, 4267, 442, 443
external waves, see waves, external
externally mixed, see size distribution, externally
mixed and particle, externally mixed
extinction coefcient, 290, 294, 295, 299, 300,
301, 315, 316, 321, 334, 335, 693, 702
aerosol absorption, 303, 3045, 315, 316, 321
aerosol scattering, 1, 2, 3045, 315, 316, 321
cloud-drop absorption, 3045, 316, 321
cloud-drop scattering, 3045, 316, 321
gas absorption, 294
gas scattering, 300
total, 299, 316
eye irritant, 366
F
Fahrenheit, Gabriel Daniel, 2
fall speed, see sedimentation, speed
family, chemical, 439, 442, 646, 648, 655, 657
aqueous, 655
odd chlorine, 439
odd hydrogen, 439
odd oxygen, 439
odd nitrogen, 439
S(IV), see S(IV)
S(VI), see S(VI)
family method of solving chemistry, 43942, 658
fast ow systems, 347
feldspars, 474
plagioclase, 474
potassium, 474
Ferrel, William, 4
Ficks rst law of diffusion, 525
eld capacity, soil moisture at, 258
ltering, 138
ne particles, see particles, ne
nite-difference method, 84, 17292, 202
analogs, 172, 173, 175, 192, 202
approximation of arbitrary order, 17781, 189
backward difference, 175, 176, 177, 179, 180,
200, 202, 551, 596
rst-order, 177, 179
fourth-order, 180
second-order, 179
third-order, 179
backward Euler, 424
linearized, 424, 425, 426
central difference, 175, 176, 177, 180, 181,
182, 189, 190, 198, 199, 202
fourth-order, 180
second-order, 176, 179
forward difference, 177
rst-order, 177
fourth-order, 180
second-order, 179
third-order, 179
forward Euler
for ODEs, 4224, 42731
for PDE, 182
forward-in-time, centered-in-space (FTCS),
182
nite-element method, 84, 1937, 542
forward Euler TaylorGalerkin, 200
Galerkin, 193
PetrovGalerkin, 193, 200
nite-volume method, 84, 169, 172, 199, 202
rst law of thermodynamics, 12, 4757, 58, 78,
603, 608
xed gases, 1416
ash photolysis, 347
fog, 5, 7, 37, 46, 335, 375, 446, 449, 462, 465,
468, 599, 602, 606, 642, 652, 682
advection, 602
evaporation, 602
frontal, 602
ice, 599
radiation, 602
steam, 602
upslope, 602
fog deposits, 599
fog drop, 447
foliage, 259
area density, 233
wind speed in, 261
force, 87
forces
apparent (inertial), 89
apparent centrifugal, see acceleration,
centripetal
Coriolis, see acceleration, Coriolis
drag, 519
electrostatic, 6, 310, 518, 638
gravitational, see acceleration, gravitational
internal, 102
794
Index
pressure-gradient, 34, 82, 90, 99101, 127
viscous, 82, 90, 92, 1014, 105, 109, 135, 507,
514, 522, 661
particle, 512
forced convection, see convection, forced
force-restore method, 259
forecast, model, 1, 5, 685, 693
forest re, see emission, forest re
formaldehyde, gas, 291, 364, 366, 367, 369, 375,
382, 384, 385, 390, 391, 444, 702, 738,
744, 751
structure, 392
formic acid, aqueous, 448, 738, 739, 740, 745,
751
formic acid, gas, 384
formyl radical, 367
forward difference, see nite-difference method,
forward difference
forward Euler, see nite-difference method,
forward Euler
forward-in-time, centered-in-space, 182
fossil-fuel combustion, 47980
four-stream approximation, 332
Fourier series, 198, 199
Fowler, Alfred, 291
y ash, 479, 480, see also ash
fractal dimension, 515, 516, 523
fractal geometry, 51416
collision correction, see collision kernel, fractal
correction
free atmosphere, 248
free convection, see convection, free
free-molecular regime, 507, 509, 512, 516
free radical, 339
freezing, 10, 37, 38, 43, 48, 7780, 270, 408,
535, 553, 598, 599, 609, 614, 6248, 632,
633, 639, 642, 643, 712
contact, 6245
evaporative, 6323
heterogeneous, 6258
homogeneous, 6258
homogeneousheterogeneous, 626, 628
freezing point
dilute water, 43
seawater, 269
freezing rain, see rain, freezing
freezing rate of drops, 627
Freon, 400
frequency, see waves, frequency of oscillation
friction, 101, 113, 118, 231, 245, 271, 475, 476,
493, 667
friction wind speed, 23049, 667
threshold, 476
front, 8, 601
cold, 601
warm, 601
frontal fog, see fog, frontal
frost, 599
frost point, see temperature, frost point
Froude number, 109, 110
frozen dew, see dew, frozen
fugitive emission, 477
full-moving structure, see size structure,
full-moving
full-stationary structure, see size structure,
full-stationary
G
Galerkin method, 193
Galileo Galilei, 2
gas
diameter, 102, 289, 447, 528, 529
dened, 8
xed, 1416
inorganic, 338, 339
list of, 71628
organic, 18, 338
variable, 1618
gas constant
dry air, 30
moist air, 33
universal, 29
water vapor, 31
gas-to-particle conversion, 10, 447
gasoline, 479
Gaussian plume, 197
Gears method, 418, 419, 4329
geometric mean diameter, 454, 455, 456
geometric size distribution, 450
geometric standard deviation, 454, 455, 456, 458,
476
geometrical regime, 308, 309
geopotential, 98, 99, 132, 146, 150, 159, 217,
221, 222, 224, 226, 685, 689, 692, 693
height, 99
magnitude, 98
vector, 99
geostrophic, 112, 113, 114, 129, 130, 132, 135,
136, 167
geostrophic adjustment, 112, 129
geostrophic wind, 82, 11114, 117, 124, 127,
128, 129, 130, 133, 136, 150, 151, 167
pressure coordinate, 133, 1501
Gibbs free energy, see energy, Gibbs free
Glasgow, 375
glaze, 599
global climate change, 6, 7, 681, 682
global model, see model, global
global warming, 294
glucose, 15
795
Index
glutaric acid, 570
glycol aldehyde, 383
gradient operator, 645
horizontal, along surface of constant
altitude, 113
pressure, 145
sigma-pressure, 153
three-dimensional
in Cartesian-altitude coordinates, 64
in spherical-altitude coordinates, 86
gradient operator conversion
altitude to pressure coordinate, 1446
altitude to sigma-altitude coordinate, 1612
altitude to sigma-pressure coordinate, 1523
pressure to sigma-pressure coordinate, 1534
gradient Richardson number, 244, 253
gradient transfer coefcient, see diffusion
coefcient, eddy
gradient transport theory, see K-theory
gradient wind, 82, 111, 11417, 136
scalar velocity
graphical calculus, 4
graphitic carbon , see black carbon
graupel, 446, 448, 610, 620, 621, 622, 624, 625,
627, 628, 633, 634, 639, 640, 665
gravitational acceleration, see acceleration,
gravitational
gravitational collection, see coagulation
gravitational force, see acceleration
gravitational potential energy, see energy,
gravitational potential
gravity, 13, 34, 35, 82, 93, 94, 95, 98, 100, 118,
119, 124, 125, 126, 127, 129, 130, 135,
136, 206, 207, 219, 222, 237, 257, 475,
610, 661, 663, 670, 679, 711, 715, see also
acceleration, gravitational
gravity waves, see waves, gravity
greenhouse effect, 25, 27, 55, 60, 76, 167, 232,
2924, 312, 326, 340, 362, 399, 402, 444,
605, 613, 619
greenhouse gas, 16
Greenwich, UK, 83
grid
irregular, 84
regular, 84
grid block, 688
grid boxes, see grid cells
grid cells, 65, 172, 173, 177, 178, 179, 182, 193,
194, 196, 202
grid domain, 683
parent, 684
progeney, 684, 690, 691
grid stretching, 84
grid volume averaging, 6874
ground-state atomic oxygen, see oxygen, atomic
ground temperature, 233, 236, 249, 255, 259,
261, 262, 263
group, see waves, group
group velocity, see waves, group velocity
growth, see condensation, deposition, dissolution
gypsum, 474
H
Haagen-Smit, Arie, 5
hail
small, 599
half-lifetime, 353
Halons, 400, 401, 405, 406, 414
H-1301, 400, 401, 405
H-1211, 400, 401, 405
H-2402, 400, 401
Hamaker constant, 513
hardwiring, 688
Hartley, John, 291
Hartley bands, 292
HCFC-22, 401
HCFC-141b, 400
HCFC-142b, 400
heat, 3, 38, 633
heat capacity, molal, see also specic heat
apparent
at innite dilution, 567
standard, 561, 7401
heat conduction equation, 256
heat-transfer coefcient (Stanton number), 238
in foliage, 261
heat transfer, dened, 38
heating rate
conductive, 256
diabatic, 47, 80
radiative, 10, 11, 320, 3324, 539, 685
helium, 16, 337, 341
hematite, 302, 448, 474, 480, 740
hemiterpene, see isoprene
Henrys law constant, 563, 656, 660
dimensionless, 584
effective, 566, 597, 637, 656, 669, 675
Henrys law, 563, 584, 597, 636, 645, 646, 652,
669, 672, 673, 674
HenyeyGreenstein function, 322
Herzberg continuum, 292
heterogeneous freezing, see freezing,
heterogeneous
heterogeneous reaction, see chemical reactions, gas
heterogeneous nucleation, see nucleation,
heterogeneous
Heun scheme, 1867
hindcast, model, 685
HITRAN, 295, 297
homogeneous equation, 169
homogeneous freezing, see freezing, homogeneous
homogeneous uid, 67
homogeneous nucleation, see nucleation,
homogeneous
796
Index
homosphere, 14, 19, 338, 339
Hooke, Robert, 1
hour angle, 319
Howard, Luke, 599
HO
x
, 397
Huggins bands, 291
hurricane, 8, 238, 682
Huygens principle, 286
hybrid size structure, 466
hybrid turbulence parameterization, 2502
hydration, 570, 571, 574, 582, 647, 655
hydraulic conductivity, see conductivity, hydraulic
hydrobromic acid, 406, 407, 444
volume mixing ratio
hydrocarbon, 338
nonmethane (NMHC), 338
oxygenated, 338
hydrochloric acid, aqueous, 400, 448, 556, 557,
565, 589, 651, 738, 739
hydrochloric acid, gas, 6, 4005, 409, 410, 413,
416, 444, 449, 471, 477, 482, 489, 531,
536, 554, 556, 557, 5638, 5735, 582,
585, 586, 587, 5917, 647, 651, 653, 675,
679, 738, 740, 746, 748
hydrochlorouorocarbons (HCFCs), 401
hydrogen, molecular, 337, 338
structure, 338
hydrogen ion, 448, 554, 555, 557, 573, 574, 576,
577, 586, 590, 596, 597, 639, 648, 739,
743, 748
hydrogen peroxide, aqueous, 360, 448, 648, 654,
655, 739, 751
hydrogen peroxide, gas, 291, 563
hydrogen sulde, 15, 341
hydrogen sulde radical, 373
hydrouorocarbons (HFCs), 400, 401, 402
HFC-134a, 402
hydrometeor, 912, 273, 304, 305, 316, 3213,
329, 333, 334, 362, 446, 508, 517, 518,
553, 598, 614, 615, 618, 6204, 628,
63142, 654, 659, 685, 686
hydrometeor particle, see particle, hydrometeor
hydroperoxy radical, gas, 360, 361, 362, 400
hydrophilic surface, 490
hydrophobic surface, 490, 536, 549
hydrostatic, 12, 34, 109, 138, 139, 140, 141, 143,
144, 150, 151, 159, 160, 163, 167, 221,
224, 610, 689
hydrostatic air pressure, see pressure, air,
hydrostatic
hydrostatic approximation, 138, 139, 140, 642
hydrostatic equation, 4, 346, 50, 57, 111, 139,
140, 143, 147, 148, 150, 154, 159, 204,
207, 2212, 224, 692
hydrostatic model, see model, hydrostatic
hydroxide ion, 557
hydroxyl radical, gas, 339, 359
concentration, 359
sources, 381
structure, 339
hydroxymethanesulfonate ion
hydroxymethanesulfonic acid (HMSA), 651
hygrometer, 2, 3, 8, 9
hypochlorous acid, 444
I
ice contact nuclei (ICN), 624, 625
ice crystal, 9, 10, 37, 40, 43, 44, 234, 277, 408,
446, 518, 54850, 598, 599, 614, 615,
626, 631, 639, 640, 665
growth, see deposition, ice
shape, 549
ice deposition nuclei (IDN), 548, 549, 615, 617,
624, 625
activation, 617
ice fog, see fog, ice
ice nuclei, see ice deposition nuclei, ice contact
nuclei
ice pellets, 599
ideal gas law, 4, see also equation of state
identity matrix, see matrix, identity
illite, 474
imaginary of refraction, 303, 304, 309
implicit equation/solution, 140, 181, 182, 184,
185, 189, 191, 195, 202, 4947
implicit time-stepping scheme, 1825
implicit loss coefcient, 425, 426, 428, 624
incompressible uid, 67, 125, 1315, 236
independent variable, 143, 144, 169, 170, 177
index of refraction
complex, 303, 309, 310, 311
imaginary, 303, 304, 309
real, 285, 301
inertia gravity waves, see waves, gravity, inertia
inertial acceleration, see acceleration
inertia Lamb waves, see waves, Lamb
inertial cascade, 235
inertial oscillation, see oscillation, inertial
inertial reference frame, 87, 88
inertial stability, see stability, inertial
inertial subrange, 235, 236
inuence functions, 248
infrared radiation, see radiation, infrared
797
Index
inhomogeneous uid, 67, 169
initial conditions, see model, initial conditions
initial-value problem, 170, 419
inorganic gas, see gas, inorganic
instability, buoyant, see stability, buoyant
instability, inertial, see stability, inertial
internal energy, see energy, internal
internally mixed, see size distribution, internally
mixed and particle, internally mixed
internal waves, see waves, internal
interpolation, 36, 57, 58, 186, 222, 240, 326,
432, 438, 442, 547, 568, 572, 573, 574,
6946
bilinear, 265, 696, 707
inverse-square, 694, 695
inversion, temperature, see temperature
inversion
ion, charge balance, see conservation, charge
iondipole interactions, 408
ionic strength, 565
iron, 650, 739, 744, see also hematite
irradiance, 260, 262, 272, 279, 280, 283, 316,
320, 324, 325, 326, 330, 331, 334, 335,
668, 693, 702, 715, 716
blackbody, 280
solar, top of atmosphere, 692, 71516
spectral, 279
irreversible reaction, 10, 553, 595, 645, 646, 647,
648, 651, 652
isentropic condensation temperature (ICT), 600,
606
isentropic coordinate, see coordinates, isentropic
isentropic surface, 52, 143, 207, 606
isobar, 100, 112, 144, 150, 151
isobaric coordinate, see coordinates, pressure
isobaric process, 49
isochoric process, 49
isomerization reaction, see chemical reactions,
gas, isomerization
isopleth, ozone, see ozone, isopleth
isoprene, 357
isothermal process, 25, 49, 125
isotropic, 279, 280, 301, 322, 323
irrotational ow, 214, 215
J
Jacobian matrix, 434
jet
nocturnal (low-level), 114
polar night, 62
jet fuel, 479, 480
jet stream
polar front, 124, 408
subtropical, 62
Julian day, 318, 319, 325
Junge layer, 478
K
k-distribution method, 297, 298
kaolinite, 474, 624
Kelvin effect, 5334
kerosene, 479
ketone, 338, 392
kinematic variable, 70, 229, 230, 235, 237, 238,
239, 240, 246, 254, 257
kinetic analysis, 346
kinetic energy, see energy, kinetic
kinetic energy, turbulent, see turbulent kinetic
energy
Kirchoff, Gustav, 277
Kirchoffs law, 277
Knudsen number
for air, 505, 506, 507, 516, 520, 662, 663, 671
for energy, 532
for trace gas, 530
KnudsenWeber term, 509
Kolmogorov scale, 236
K ohler equation, 5368
Koschmieder equation, 299, 315
krypton, 16
K-theory, 72, 106
Kuo schemes, 613
Kutta, Wilhelm, 4
L
Lagrangian, 63, 192, 201, 203, 204, 463, 466,
682
Lagrangian trajectory model, 682
Lamarck, Jean Baptiste, 598
Lambert conformal projection, see coordinates,
Lambert conformal
Lamb waves, see waves, Lamb
laminar ow, 69
land breeze, 377
land-use data, 692
lapse rate, 25, 50, 53, 59, 603, 605, 606, 607, 643
environmental, 35, 53, 54, 604, 607
dry adiabatic, 50
actual temperature, 50
virtual temperature, 50, 55
pseudoadiabatic, 6023
large-eddy simulation, see model, large-eddy
simulation
latent heat, 3840
of evaporation, 38
of melting, 38
of sublimation, 38
latent heat ux, 258
latent heat release/absorption, 47
latent heat transport, 274
latitude, 25, 26, 83
laughing gas, see nitrous oxide
lead suboxide, 448, 740
leaf area index, 261, 670
leaf water content, 263
798
Index
leapfrog scheme, see time-stepping schemes,
leapfrog
letovicite, see triammonium bisulfate
level of free convection (LFC), 600, 612
level of neutral buoyancy (LNB), 612
Lewis, Gilbert, 275, 336, 338
Lewis structure, 338
Lewis symbols, 336
lifetime, chemical, 353
e-folding, 353
half, 353
overall, 353
lifting condensation level (LCL), 600, 601, 605,
606, 607, 610, 612, 643
light, speed of, 285
light processes, 28390
lightning, 10, 336, 478, 479, 480, 63842
limited-area model, see model, limited-area
limonene, 388
line-by-line data, 2938
line distribution, 295
air-broadened pressure shift, 296
central wavenumber, 296
halfwidth
air-broadened, 296
pressure-broadened, 296
self-broadened, 296
intensity, 296
linear differential equation, 169
linearized equation, 424, 426
liquid water content, see particle, liquid water
content
loam, 257
loamy sand, 257
local derivative, 63, 78, 138, 141
locally one-dimensional (LOD) procedure, 171
logarithmic wind prole, see wind, speed
lognormal distribution, see size distribution,
lognormal
London, 375
London-type smog, see smog
longitude, 25, 83, 683
longitudinal wave, see waves, longitudinal
Lorentz prole, 295
Lorenz grid, 205
Los Angeles, 5, 377
low level jet, see jet, nocturnal
low-pressure impactor, 455
low-pressure system, 14
lubricating oil, 480
lumped species method, 391, 393
M
M, 27, 343, 346
magma, 477
magnesium carbonate, 302, 448, 739, 741
magnesium chloride, 448, 739
magnesium ion, 448, 479, 482, 554, 739
magnesium nitrate, 448, 739
magnesium sulfate, 448, 739
malonic acid, 570
manganese, 650, 739, 744
MarshallPalmer distribution, 45960, 468
mass
one air molecule, 18, 36
one particle, 519
one trace gas molecule, 410
mass absorption coefcient, 295
mass accommodation coefcient, 411, 530, 531
mass concentration, 12, 449, 453, 455, 457, 458,
459, 468, 469, 494, 552
mass conservation, 420
mass density, see density
mass ux iteration (MFI) method, see equilibrium,
model, solution methods, MFI
mass mixing ratio, see mixing ratio, mass
material time derivative, see total derivative
mass transfer coefcient of gas, 655
matrix
identity, 434
Jacobian, 434
order, 434, 435
predictor, 433, 434, 435
reordering, 436, 464
solution methods
decomposition/backsubstitution, 183, 189,
191, 433, 436, 437, 438, 543, 658
Gaussian elimination, 436
sparse, techniques, 432, 434, 436, 543, 659,
688
tridiagonal, 183
Matsuno scheme, 186
matter, dened, 37
Maxwell, James Clerk, 526
MaxwellGarnett mixing rule, 310
Maxwellian distribution, 18
mean anomaly of the Sun, 318
mean free path
of air molecule, 505, 506, 507, 661, 662, 663
of a particle, 509
of a trace gas, 530
of energy (thermal), 532
mean longitude of the Sun, 318
mechanical shear, see shear, mechanical
mechanical turbulence, see turbulence, mechanical
median freezing temperature
equilibrium, 626
nonequilibrium, 628
melting, 10, 80
ice crystal, 6334
melting point
ice, 268, 270, 633, 634, 712, 713
Mercator projection, see coordinates, Mercator
meridian, see also longitude
prime, 83
mesopause, 28
799
Index
mesophyll, 668, 669, 679
mesoscale motions, 8
mesosphere, 19, 28, 57, 282
meteoric debris, 481
Meteorologica, 3
meteorological model, 7
meteorological range, see visibility
meteorology
dynamical, 4, 6
history of, 15, 6
methacrolein, 389
methane, 7, 12, 14, 16, 17, 292, 294, 311, 338,
342, 354, 363, 376, 379, 383, 386, 393,
398, 399, 478, 479, 701
structure, 342
methanesulfenic acid, 372
methanesulfonic acid (MSA), 371
methanethiol, 369, 373
methanethiolate peroxy radical, 371
methanethiolate radical, 370
methanogen, 17
methanol, gas, 390
method of fractional steps, 171
methoxy radical, 364
methyl bromide, 400, 401, 405, 406, 444
methyl chloride, 292, 294, 400, 401, 403, 404,
444
methyl chloroform, 400, 401, 444
methyl criegee biradical, 385
methylene glycol, 650
methylethylketone, 392
methyl hydroperoxide, gas, 364
methylperoxy radical, gas, 364
methyl radical, 343
methylvinylketone, 389
Meuse Valley, Belgium, 375
microscale motions, 8
Midgley, Thomas, 399
Mie regime, 307, 308
absorption in, 308
scattering in, 307
mineral, 266, 473, 474
mist, 3, 10, 401, 599
mixed layer, ocean, 268
mixing length, 246, 252, 506
free-atmospheric, 251
mixing ratio, mass
dry air, 31, 32, 33, 45, 48, 58, 59, 66, 72, 74,
81, 144, 148, 153, 212, 223, 602, 605,
606, 608, 610
water vapor, 32
saturation over liquid, 45
mixing ratio, moist-air mass, 32
water vapor, see specic humidity
mixing ratio, volume, 14, 16, 346, 444
dened, 31, 32, 58, 334, 415, 444
mixing rule, activitity, 310, 311, 334, 563, 569,
595, see also Bromleys method, ZSR
method
mixing rule, optical, 309, 311
Bruggeman, 311
MaxwellGarnett, 310
inverted, 310
volume average, 309
volume average dielectric constant, 310
mixing state, see particle, mixing state
mobile source emission, 380
mobility, 519, 639
mobility radius, 515, 516, 523
model, 681707
air quality, 6
air-pollution-weather-climate, 8
atmospheric, 65
boundary conditions, 83, 6902
limited-area, 183, 684
periodic, 182, 184, 185, 188, 197, 201, 202,
203, 210, 215, 226, 690
virtual, 196, 690
box, see model, zero-dimensional
coding of, 6878
computer architecture for, 687
computer, atmospheric, 6
dened, 6
design of, 681707
dimensions of, 686
direct numerical simulation (DNS), 105, 236
global, 683
hemispheric, 683
hydrostatic, 138140, 143, 167, 684
initial conditions, 68990
input data, 6923
interpolation for, see interpolation
Lagrangian trajectory, 682
large-eddy simulation (LES), 236
limited-area, 183, 196, 209, 212, 218, 683,
684, 700
meteorological, 7
nested, 84, 209, 437, 438, 683, 684, 687, 690,
691
one-way, 684
two-way, 684
800
Index
nonhydrostatic, 13843
anelastic, 139, 14050
elastic, 139, 140, 159, 160, 165, 166
ocean, 269
one-dimensional, 682
optimization of, see computer, optimization
parcel, see model, zero-dimensional
photochemical, 7
processes in, 685
Reynolds-averaged, 68
sensitivity tests, 700
simulations, 699707
soil, 259
statistics, 6979
three-dimensional, 683
time interval, see time interval
time step, see time step
two-dimensional, 683
variables in, 6856
zero-dimensional, 682
modied gamma distribution, 461
moist adiabatic, see adiabatic process, moist
moist convective adjustment scheme, 613
moisture potential, see soil, moisture potential
molal enthalpy of formation, 561
molality, 534, 559, 56284, 597, 617, 636, 645,
646, 6727, 749
molarity, 555, 556, 557, 597, 669
mole concentration
gas, 541
saturation, 544, 584, 617
particle, 544
mole conservation, see conservation, mole
molecular diffusion, see diffusion, molecular
molecular mass, see mass
molecular-scale motion, 8
molecular weight
dry air, 30, 32, 712
moist air, 34
trace gas, 30, 347, 520
water vapor, 31
molozonide, ethane, see ethene molozonide
momentum equation, 11, 61, 63, 80, 82136,
13850, 15670, 21119, 225, 229, 322,
598, 610, 611, 612, 642, 701
aloft along straight isobars, 11213
aloft around curved isobars, 11417
horizontal
sigma-pressure coordinate, 159, 160, 21421
quasigeostrophic, 133
surface along straight isobars, 11314
surface around curved isobars
three-dimensional
Cartesian-altitude coordinates, 110
spherical-altitude coordinates, 111
vector form, 108
vertical
in a cloud, 61011
MoninObukhov length, 240, 241, 271, 667
MoninObukhov similarity theory, 250
monodisperse size distribution, 498, 499, 523
monomer size distribution, 495, 497, 498
monoterpene, 388
monotonic, 199, 200
montmorillonite, 624
Montreal Protocol, 414
Mount Pinatubo, 398
mountain lee waves, 127
moving-center structure, 4667
multiprocessor machine, 686, 687
multistep implicit-explicit (MIE) solution, 418,
42732, 442, 443, 689
N
nanoparticles, 512, 516, 523
natural gas, 479
neon, 16
nested model, see model, nested
neutral stratication, see stability, buoyant
Newtons law of universal gravitation, 96
Newtons second law of motion, see also
momentum equation, 87111
nighttime chemistry, 3623, 400
nimbostratus, see cloud types, nimbostratus
nitrate ion, 15, 302, 448, 556, 566, 567, 589
nitrate radical, gas, 292, 343, 351, 355, 3608,
3803, 386, 389, 393, 444, 448, 482, 531,
534, 554, 558, 563, 564, 566, 573, 574,
577, 58797, 651, 653, 654, 675, 702, 748
nitrated aromatics, 302
nitric acid, aqueous, 362, 448, 563, 564, 577,
589, 596, 651
nitric acid, gas, 291, 556, 562, 563, 564, 566,
573, 575, 577, 596, 597, 738, 739, 740,
741, 746, 751
nitric acid dihydrate (NAD), 408
nitric acid trihydrate (NAT), 408
nitric oxide, 339
production by lightning, 642
structure, 339
nitrication, 15
nitrifying bacteria, 15
nitrite ion, 15
nitrobenzene, 304
nitrocresol, 388
801
Index
nitrogen, molecular, 5, 12, 14, 15, 28, 291, 292,
300, 336, 339, 340, 343, 346, 362, 365,
366, 375, 376, 378, 380, 393, 396, 409,
424, 425, 477, 478, 638, 682
concentration, 346
structure, 339
nitrogen dioxide, gas, 18, 290, 291, 292, 299,
340, 342, 351, 35779, 38396, 401, 405,
407, 409, 410, 413, 415, 416, 4204, 427,
434, 443, 444
structure, 340
nitrogen xation, 15
nitrosofying bacteria, 15
nitrous acid, gas, 291, 359, 360, 361, 379, 380,
738, 739, 751
nitrous oxide, 7, 15, 294, 396, 478
nocturnal boundary layer, 24
nocturnal jet, 114
nodes, see grid cells
nondivergent, ow, 67, 133, 135, 142, 214, 215
nonequilibrium particle growth, 577, 58394, see
also dissolution
of solids, 591
nonhydrostatic air pressure, 12, 141
nonhydrostatic model, see model, nonhydrostatic
nonhydrostatic pressure, 142
noninertial reference frame, 87, 110
nonmethane hydrocarbons, see hydrocarbon,
nonmethane
nonmethane organic carbon (NMOC)
nonoscillatory, 199, 200, 588
normalized bias, 697, 698, 707
normalized gross error (NGE), 420, 697, 698,
699, 702
location-specic, 697
time-specifc, 697
NO
x
(nitrogen oxides), 376
nucleation, 4, 14, 233, 267, 281, 325, 369, 372,
373, 378, 385, 390, 475, 48492, 525,
537, 557, 570, 578, 601, 636, 661
binary, 484, 488
classical theory, see classical nucleation theory
dened, 10, 460, 470, 482, 484, 485
heterogeneous, 74, 76, 484, 48991
homogeneous, 4849, 4912, 633
solved with condensation, 5457
homomolecular, 484, 486
ternary, 484, 489
nucleation mode, 457, 460
composition, 482
nucleation scavenging, see rainout
nudging, 690
number concentration
cloud drops, 447
dry air, 30
gas, 63, 447
moist air, 30
particles, 75, 453
water vapor, 30
numerical diffusion, 175, 197, 200, 203, 462,
464, 466, 467, 468, 542
numerical dispersion, 175
numerical solution, 421
numerical stability, 53, 55, 174, 222, 586, 605,
606
conditional stability, 174, 181, 182, 186, 187,
191, 223
unconditional instability, 174, 181, 183, 189,
191, 202, 425, 426, 542, 544, 545, 585,
593, 620, 637
unconditional stability, 174, 182, 186, 187,
191, 222, 419
O
oblate spheroid, see Earth, shape
obliquity of the ecliptic, 318
ocean, 16, 17, 24, 67, 83, 84, 93, 114, 119, 232,
233, 238, 26875, 370, 373, 377, 458,
4703, 479, 4913, 555, 566, 570, 602,
612, 661, 6729, 682, 685, 692
composition, 470, 471
mixed layer, 268
model, see model, ocean
octet, 338
oil, see petroleum
olens, 392
operator splitting, 1701
optical depth, 31617
optical focusing effect, 311
order of approximation, 169, 174
order of closure, 237
zero, 237
rst, 237
order of matrix, see matrix, order
ordinary differential equation, 4, 6, 7, 11, 169,
336, 352, 355, 415, 418, 419, 421, 442,
542, 544, 614, 645
solutions to, 41821, 6549
organic carbon, nonmethane, see nonmethane
organic carbon
organic carbon, see organic matter
organic gas, see gas, organic
organic matter, 15, 16, 290, 448, 449, 458, 473,
474, 477, 478, 479, 480, 536, 549, 693,
see also soot; organic carbon, particulate
ultraviolet absorption by, 302
orographic cloud, see cloud types, orographic
orographic uplifting, 601, 642
oscillations, 57, 118, 127, 128, 129, 130, 131,
197, 199, 200, 205
802
Index
buoyancy, 57, 124
inertial, 127
pressure, 122
oxalic acid, 570
oxygen, atomic, 26, 291, 292, 342, 443
oxygen, excited atomic, 26, 27, 292, 354, 359,
369, 396
oxygen, molecular, aqueous
oxygen, molecular, gas, 12, 14, 15, 16, 18, 26, 27,
28, 282, 291, 292, 300, 336, 3389, 340,
342, 343, 346, 359, 375, 376, 394, 395,
396, 414, 424, 425, 439, 442, 474, 478,
480, 483, 570, 638, 650, 652, 668, 739
concentration, 346
ozone formation from, 3946
structure, 338
ozone, aqueous
ozone, gas, 5, 6, 7, 8, 12, 14, 16, 18, 26, 27, 28,
32, 58, 81, 124, 282, 291, 292, 294, 299,
300, 33342, 35768, 376415, 423, 425,
439, 580, 581, 648, 649, 652, 669, 672,
6805, 693, 701, 702, 707, 739, 751
absorption of radiation by, 291
catalytic destruction cycles, 396, 397, 404, 406,
412
column abundance, 394, 395, 398, 407, 414,
415
dent, Arctic, 408
formation, 268, 357415
health effects, 340
hole, Antarctic, 6, 7, 8, 398, 399, 40715
isopleth, 376
layer, 26, 397
effects of bromine on, 398407
effects of carbon on, 398
effects of chlorine on, 398407
effects of hydrogen on, 397
effects of nitrogen on, 3967
odor, 340
peak density, 27
peak mixing ratio, 27, 393
peak number concentration, 393
recovery, 41415
reduction
Antarctic, see ozone, hole, Antarctic
global, 32, 399
regeneration, 414
structure, 340
volume mixing ratio, 18, 340, 358, 393, 394,
415
P
Pacic high, 378
paired peak accuracy, 699
parafn, 391
parallel processor, 687
parameterized equation, 7
partial derivative, 86, 435, 543
partial differential equation, 5, 6, 7, 11, 169202,
420
partial pressure, 30
dry air, 30, see also pressure, air, dry
trace gas, 30
water vapor, 30, 31
particle
aerosol, 8
components, 557, 73840
composition, 4479
dened, 9
density, 453
diameter, 447
externally mixed, 309, 447, 494, 502, 503, 523
ne, 458
ow regimes, 5057
hydrometeor, 9
internally mixed, 309, 447, 494, 500, 505, 523
liquid water content
aerosol, 573, 645
cloud, 645
fog, 447
precipitation, 34, 255, 258, 262, 266, 460,
461, 570, 580, 585, 591, 593
mass, 519
mass concentration, 12, 449, 453, 455, 457,
458, 459, 468, 469, 494, 552
mixing state, 30912
number concentration, 75, 453
primary, 447
radius, see radius, particle
scattering, see scattering, particle
secondary, 447
soil dust, see soil dust
surface area concentration, 410, 453
thermal conductivity, 520
volcanic, 398
volume
component, 309
single-particle, 75, 309, 451
volume concentration
individual component, 75, 453
in size bin, 75, 453
particle rebound charging, 639
partitioning ratios, 440, 441, 442
parts per million
by mass, 32
by volume, 17, 32
pathlength
mass, 295, 297
molecular, 295
pentane, 381
peruorocarbons (PFCs), 400, 402
peruoroethane, 402
803
Index
perihelion, 319
period of oscillation, 57, 124
periodic boundary conditions, see model,
boundary conditions, periodic
periodic table, 336
permeability, coefcient of, 257
permittivity, 310
of free space, 638, 713
peroxyacetyl nitrate (PAN), gas, 291, 342, 739,
751
structure, 342
peroxynitric acid, gas, 291, 360, 388, 397, 738,
739
photolyis coefcient, 352
petroleum (oil), 386, 479, 480
pH, 555, 556, 557, 585, 591, 597, 647, 648, 649,
650, 651, 652, 655, 658, 659, 660, 672,
675, 677, 678, 679
acidic rainwater, 556
natural rainwater, 557
phase speed, see waves, phase speed
phenols, 302
photochemical model, see model, photochemical
photochemical smog, see smog, photochemical
photochemistry, see chemical reactions,
photochemical
photodissociation, see photolysis
photolysis, 26, 3424
dened, 9
organic gas breakdown by, 3801
rate coefcient, 279, 3324, 344, 351, 442, 598
list of, 72838
reactions, aqueous, 7436
reactions, gas, 342, 351
photon, 15, 26, 273, 275, 276, 283, 284, 320,
333, 342, 351
photoprocess, see photolysis
photosphere, see sun, photosphere
photostationary state relationship, 358
photosynthesis, 15, 17, 338, 668
anoxygenic, 15
green-plant, 15, 16, 338
physical-property number, 664
phytoplankton, 369
pinene, 388
Pitzers method, 565, 566, 597
Pitzer parameters, 565, 566
Planck, Max, 275
Plancks constant, 275, 333, 711
Plancks law, 273, 276, 278, 334
planetary albedo, see albedo
planetary scale motions, 8
plant debris, 481
Poisson partial differential equation, 142
Poissons equation, 51
polar stratospheric clouds, see clouds, polar
stratospheric
polar vortex, 408, 413, 682
poles, 4, 25, 26, 42, 53, 62, 83, 85, 93, 95, 98,
209, 210, 225, 274, 275, 550, 598, 690
pollen, 481
polycyclic aromatic hydrocarbons (PAHs), 302,
478, 480
polydisperse size distribution, see size distribution,
polydisperse
positive-denite numerical scheme, 419, 420, 442,
636, 637
potassium bicarbonate, 448
potassium bisulfate, 448, 740
potassium carbonate, 448, 739
potassium chloride, 448, 740, 741
potassium ion, 16, 336, 448, 449, 477, 479, 482,
534, 585, 739
potassium nitrate, 448, 740
potassium sulfate, 448, 740
potential energy, see energy, potential
potential temperature, 503
dry air, 50, 51
equivalent, 604
gradient, dimensionless, 242
moist air, 51
scale, 241
virtual, 52
potential virtual temperature, 51
prole in a cloud, 609
Prandtl number, 532, 668
turbulent, 242
precipitation, hydrometeor, 10, 554
precipitation, solid, 5578, 5745
reactions, 5578
predictor matrix, 433, 434, 435
pressure, air, 2, 11, 12, 13, 14, 30, 32, 34, 35, 42,
45, 49, 58, 59, 60, 61, 136, 140, 151, 156,
204, 206, 207, 235, 272, 335, 344, 346,
520, 529, 550, 551, 606, 607, 629, 630,
634, 643, 663, 689, 693, 701, 714
dry, 30
hydrostatic, 12
model
column, 156, 208
surface, 206
top, 206
moist-air (total), 30
nonhydrostatic, 12, 141
standard sea-level, 14, 29
vertical prole, 347, 71415
pressure coordinate, see coordinates, pressure
pressure-dependent reaction, see chemical
reactions, gas
pressure gradient, 62
pressure-gradient force, see forces,
pressure-gradient
804
Index
pressure oscillation, 122
pressure systems
low, 14
high, 22, 114
Pacic, 378
prevailing visibility, see visibility, prevailing
primary particle, 447
primary pollutant, 376
primitive equations, 111
principal valence, 337
probability interval, 297, 298
probable speed of air molecule, 18
prognostic equation, 143
for column pressure, 156, 208
propane, 365
propene, 365
structure, 392
propionaldehyde, 392
propoxy radical, 366
propyl nitrate, 366
propyl nitrite, 366
propylperoxy radical, 366
pseudoadiabatic, see adiabatic process, lapse rate
pseudo-rst-order rate coefcient, 344
pseudo-second-order rate coefcient, 344
pseudospectral method, 1979
psychrometer, 3
Public Health Act of 1848, 6
pulse radiolysis, 347
Q
quadrature approximation, 328, 330
quantum number, 275
quantum yield, 333, 351
quartz, see silicon dioxide
quasigeostrophic momentum equation, 133
quasigeostrophic potential vorticity equation, 134
quasistationary structure, 183, 465
quasi-steady-state approximation (QSSA), 4267
R
radiance, 276, 277, 278, 279, 280, 294, 295, 297,
303, 320, 328, 330, 331, 332, 333, 334,
335, 539
blackbody, 279
downward, 328
spectral, 276
upward, 328
radiation, 21, 37, 273334, 369, 746, see also
radiance, irradiance, actinic ux
dened, 21
diffuse, 320
direct, 320, 321
gamma, 280
infrared, 7, 10, 17, 19, 21, 22, 23, 25, 26, 258,
260, 272, 273, 274, 276, 277, 281, 283,
290, 292, 293, 294, 295, 297, 301, 320,
321, 328, 331, 332, 334, 402, 538, 602,
690, 693, 706
solar (near), 281
thermal (far), 16, 281
solar, 10, 21, 22, 25, 28, 74, 255, 272, 273,
274, 280, 282, 291, 292, 293, 301, 302,
320, 321, 324, 330, 331, 333, 334, 353,
538, 539, 692, 699, 701, 702, 706
spectrum, 280, 281, 283, 291, 299, 321
spectrum, 275
visible, 281
ultraviolet, 7, 26, 27, 273, 281, 292, 302, 312,
313, 330, 334, 393, 394, 395, 402, 478,
693
far, 281
near, 281
UV-A, 282
UV-B, 282
UV-C, 282
X, 280
radiation fog, see fog, radiation
radiative cooling effect, 5389
radiative cooling/heating, 47
rate, 80, 539
at drop surface, 526
radiative transfer equation, 11, 273, 32034
solutions to, 32632
radius, particle, 453
aerodynamic, 515
area-equivalent, 515, 516, 523
collision, 515, 516, 517
fractal, 515, 516, 517, 523
mobility, 515, 516, 523
Stokes, 515
volume-equivalent, 515, 516, 523
radius of curvature, 115, 485
radius of inuence, 694
rain, 1, 599, 669
freezing, 599
supercooled, 599
rainbow, 282, 286
primary, 286
secondary, 286
raindrop, 9, 17, 37, 74, 286, 309, 334, 335, 446,
447, 449, 459, 460, 468, 494, 510, 511,
602, 622, 663, 664
rainfall rate, 460
rainout, aerosol, 10, 446, 458, 598, 634, 635,
636, 642
Raoults law, 5356, 537, 570
rate coefcient, chemical, see chemical rate
coefcient
Rayleigh limit for disruption, 520
Rayleigh regime, 306
805
Index
Rayleigh scattering, see scattering, Reyleigh
reaction probability, 40912
reaction rate, see chemical reactions, gas
reaction rate coefcient, see chemical rate
coefcients
reactive organic gas (ROG), 338, 3801
reactivity, 381
real index of refraction, see index of refraction,
real
receptor region, 378
reference height, 36, 238, 260, 261, 271, 272,
611, 666, 667, 680
reection, 234, 2834, 478
critical angle, 288
total internal, 288, 289
reectivity, see albedo
refraction, 2845
refractive index, see index of refraction
relative humidity, 32
relaxation at boundaries, 690, 691
residual, 24, 114, 193, 701
residual layer, 24
resistance, 665, 666, 668, 669, 679, 680
aerodynamic, 260, 666, 667, 675
buoyant convection, 668
canopy, 261, 668
cuticle, 668, 669
data, 74951
exposed surfaces, 668
mesophyll, 668, 669
minimum, 668
molecular diffusion, 666, 667, 675
soil and leaf litter, 668, 670
stomata, 261, 668
surface, 666
over ocean, 675
total, 666, 668, 672, 679, 680, 749, 751
resonance structure, 339
respiration, cellular, 16
resuspension, see emission, resuspension
Reynolds, Osborne, 68
Reynolds assumption, 69
Reynolds-averaged model, 68
Reynolds averaging, 68
Reynolds decomposition, 69, 89
Reynolds number
friction, 476
particle, 507
Reynolds stress, 102, 103, 105, 228, 229, 230,
257, 473, 475
rhumb line, 83
Richardson, Lewis Fry, 4, 5, 204, 244, 418
Richardson number
bulk, 243, 244
critical, 245
gradient, 244, 253
termination, 245
rime, 599
roads, 11, 74, 228, 265, 268, 482
road temperature, see temperature, road
Robinson, John T. R., 2
roofs, 11, 228, 265, 266, 267, 268, 271
roof temperature, see temperature, roof
root-mean-square speed of air molecule, 18
Rossby number, 109, 110
Rossby radius of deformation, 129, 130, 134, 135
Rossby wave, see waves, Rossby
roughness length, see surface roughness length
Runge, Carle, 4
RungeKutta method, 1878
S
S(IV) family, 64652
oxidation
aqueous-phase, 64752
gas-phase, 647
S(VI) family, 64652
saltation, 475, 476, 477
layer, 475
sand, 257
sandy clay, 257
sandy clay loam, 257
sandy loam, 257
Santorre, Santorio
saturation ratio, 484, 487, 488, 489, 493, 538,
616, 617
ambient, 536, 537
critical, 536, 537, 538, 616, 617
equilibrium, 536, 537, 539, 544, 550, 551, 615,
616, 655
saturation vapor pressure, see vapor pressure,
saturation
scalar, dened, 61
scalar operator, 64
scalar processor, 686, 687
scalar velocity, see wind scalar velocity
scale height, 36
scattering, 10, 287, 288, 289, 290, 301, 308, 309,
334, 340, 446, 508
back, see backscattering
cross section, see cross section, scattering
efciency, particle
forward, 308
total, 305, 308
gas, see scattering, Rayleigh
geometric, 308
isotropic, 322
Mie, 307
multiple, 320
particle, 287313
phase function, 321, 330
Rayleigh (gas), 289301, 313, 322
selective, 289
side, 287
single, 320
Tyndall, 306
806
Index
scavenger, 359, 361
Schmidt number
gas, 531, 676
particle, 510
SchumannRunge system, 292
Schwartzchilds equation, 328
scrubber, 6
sea breeze, 62, 377
sea ice, 11, 16, 228, 235, 269, 270, 271
density, 270
emissivity, 269
melting point, see melting point, ice
thermal conductivity, 269
temperature, 26870
thickness, 269
sea spray, 470, 503
acidication, 470, 566
emission, see emission, sea spray
seawater, see ocean
secondary particle, see particle, secondary
secondary pollutant, 376
secondary rainbow, see rainbow, secondary
sedimentation, 10, 37, 76, 77, 80, 231, 233, 235,
269, 271, 317, 375, 446, 470, 476, 493,
550, 553, 6615, 666, 6702, 679, 683,
692
speed, 507, 511, 519, 661, 664
semiimplicit equation/solution, 140, 181, 184,
419, 420, 424, 496, 497, 498, 499, 500,
501, 522, 524, 620
semi-Lagrangian method, 169, 1912, 202, 542
sensible heat ux, see turbulent ux, sensible
heat
series expansion method, 169, 172, 1929, 200,
202
sesquiterpenes, 388
shared memory machine, 686, 687
shear, 102
dimensionless, 240, 246, 667
mechanical, 73, see also turbulence, mechanical
production rate of, 253
shearing stress, see stress, shearing
side-scattering, 287
sigma-altitude coordinate, see coordinates,
sigma-altitude
sigma-pressure coordinate, see coordinates,
sigma-pressure
silicon dioxide, 302, 448, 474, 480, 740
silt loam, 257
silty clay, 257
silty clay loam, 257
similarity relationship, 240, 242, 246
similarity theory, 240, 247, 250
MoninObukhov, 250
Surface-layer, 240
single-processor machine, 686, 687
single-scattering albedo, 327
singularity, 83, 690
size bin, 449
diameter width of, 452
number in a distribution, 451
volume-width of, 452
size distribution, 446, 44963, 46870, 475, 477,
494510, 516, 518, 523, 547, 592, 614,
619, 624, 625, 631, 632, 636, 637, 644,
686
continuous, 449, 45461
discrete, 44954
evolution of, 4627
externally mixed, 500, 504
geometric, 450
internally mixed, 502
lognormal, 4547, 459, 460
MarshallPalmer, 45960, 468
modied gamma, 461
monodisperse, 498, 499, 523
monomer, 495, 497, 498
multiple, 5005
polydisperse, 498
volume ratio, 450
size grid, 446, 457
size parameter, 305, 539
size structure, 446, 462, 463, 464, 465, 466, 467,
468, 497, 499, 542, 546, 548, 549
full-moving, 4635
full-stationary, 4623
hybrid, 466
moving-center, 4667
quasistationary, 183, 465
slip ow, 507, 509, 662, 663
sky, color, 289
smectite, 474
smog
dened, 5, 6, 7, 341, 342, 368, 375, 376, 377,
378, 379, 381, 383, 393, 394, 682
London-type, 5, 375
photochemical, 5, 341, 37593
urban, 7
Smoluchowskis solution, 499, 523, 524
SMVGEAR II, 181, 192, 419, 451, 496, 5423,
689
Snells law, 284
snow, 234, 599
density, 268
depth, 268
emissivity, 268
temperature, 26870
thermal conductivity, 268, 713
snowfall rate, 268
snowake, 9, 44, 446
snow grains, 599
snow pellets, 599
sodium bisulfate, 448, 740
sodium chloride, 302, 303, 448, 458, 535, 538,
574, 575, 582, 592, 596, 653, 740, 741,
747, 749
807
Index
sodium ion, 448, 479, 482, 534, 562, 573, 596,
739
sodium nitrate, 302, 448, 740
sodium sulfate, 448, 740
soil, 254, 255, 257, 259, 277, 284, 302, 473, 477,
482, 503, 514
density, 256
model, 259
moisture (water content), 25565
in vegetated soil, 25965
moisture potential, 256
organic matter, 473
specic heat, see specic heat, soil
temperature, 19, 256, 264
in vegetated soil, 25965
soil dust, 302, 474
emission, see emission, soil dust
solar constant, 325, 326, 711, 716
solar radiation, see radiation, solar
solar spectrum, 281
solar zenith angle, see zenith angle, solar
solid angle, 276, 277, 278, 279
solid precipitation, see precipitation, solid
solstice, 319
solubility, 553, 554, 567, 574, 677
solubility product, 564, 677
solute, 448, 471, 527, 5349, 551, 553, 559,
56374, 578, 589, 5957, 616, 637, 653,
673
solute effect, see Raoults law
solution, 448, 535, 553, 559, 570, 673
solvation, 570
solvent, 401, 448, 535, 553, 559, 570, 673
soot, 290, 302, 304, 306, 307, 308, 311, 312,
315, 324, 335, 458, 478, 479, 480, 503,
514, 515, 523, 536, see also black carbon,
organic matter
sound, speed of, 123
sounding, 690, 701
source region, 378
SO
x
(sulfur oxides), 378
sparse-matrix techniques, see matrix, sparse,
techniques
specic heat, 1, 1922, 23, 24, 38, 55, 114,
255, 261, 292, 313, 340, 357, 358, 3602,
368, 377, 380, 381, 386, 602, 667, 669,
680, 690, 706, see also heat capacity,
molal
dry air at constant pressure, 20
dry air at constant volume, 48
ice, 38
liquid water, 19, 20, 38, 269
moist air at constant pressure, 49
moist air at constant volume, 48
soil, 19, 256
clay, 20, 22, 255, 264, 265, 267, 475, 477
dry sand, 20
soilwaterair mixture, 255, 256
water vapor at constant pressure, 38, 49
water vapor at constant volume, 48
specic humidity
foliage air, 261
leaf stomata, 261
saturation, 260
vertical prole of, 249
specic humidity gradient, dimensionless, 246
specic volume of air, 47
spectral method, 193, 197
spectral transform method, 199
speed of light, 285
speed of sound, 123
spherical coordinates, see coordinates, spherical
spherical harmonics, 199
spherule, 312, 478, 515
spores, 481
spray cans, 401
spume droplets, 470
stability, buoyant, 6046
absolute stabililty, 185, 604, 605
conditional instability, 604, 605
criteria
saturated air, 6046
unsaturated air, 537
neutral stratication, 23, 537, 240, 248, 6047
stable stratication, 24, 53, 54
unstable stratication, 24, 53, 54, 55, 57, 240,
248, 6047
stability, inertial, 128
instability, 127, 128
neutrality, 128
stability, 127
stability, numerical, see numerical stability
stable boundary layer, see boundary layer, stable
standard atmosphere, 35
Stanton number, see heat-transfer coefcient
static reaction systems, 347
stationary source emission, 380
steady ow, 69
steam engine, 6
steam fog, see fog, steam
StefanBoltzmann constant, 283
StefanBoltzmann law, 283
steradian, 276, 277, 278, 279, 280, 333
stereographic projection, see coordinates,
stereographic
steric factor, 347
sticking coefcient, see accommodation coefcient
stiff system of equations, 3535
stiffness predictor, 438
stoichiometric coefcient, 554, 555, 563, 565
Stokes, George, 507
StokesEinstein equation, 509
Stokes ow, 507, 509, 661, 662, 663
Stokes law, 661, 662
Stokes number, 511
Stokes radius, 515
808
Index
Stokes terminal fall speed, 662, 664
stomata, 255, 261, 262, 668, 669, 679
stratocumulus clouds, see cloud types,
stratocumulus
stratopause, 25
stratosphere, 18, 19, 268
stratus, see cloud types, stratus
streamfunction equation, 214
street canyon, 82
stress
Reynolds, 228, 229, 230, 257
shearing (viscous), 102
structures, chemical, 74, 238, 282, 339, 597,
71628
Strutt, Robert John, 291
subcloud layer, see boundary layer, subcloud
subgrid scale disturbances, 6874, 2357
sublimation, 10, 16, 37, 38, 40, 43, 48, 74, 75,
76, 77, 78, 80, 268, 270, 447, 448,
52550, 553, 556, 570, 598, 599, 610,
614, 615, 618, 628, 629, 631, 636, 642,
645, 646, 649, 653, 669
subsolar point, 317, 318
succinic acid, 570
sucrose, 570, 571
sulfate aerosol, 308, 315, 446, 454, 458, 474,
477, 479, 480, 502, 503, 556, 564, 568,
589, 646, 647, 648, 649, 652, 659, 693,
701, 702, 707
sulfate ion, 448, 482, 534, 535, 562, 589, 646,
649, 739
sulte ion, 448, 646, 657, 739
sulfur, elemental, 15
sulfur dioxide, aqueous, 646
sulfur dioxide, gas, 341
structure, 341
sulfur hexauoride, 400, 402
sulfur monoxide, 371
sulfur photochemistry, 36975
sulfur trioxide, 375
sulfuric acid, aqueous, 302, 303, 369, 374, 448,
457, 478, 554, 556, 586, 589, 646, 653,
739, 740
sulfuric acid, gas, 369
sulfurous acid, 448, 449, 646
Sun, 276, 280, 281, 283, 284, 286, 287, 289, 290,
316, 317, 318, 319, 324, 325, 340, 412,
586, 600, 692, 712, 713
color, 289, 290
ecliptic longitude, 317
luminosity, 324
mean anomaly, 318
mean longitude, 318
photosphere, 280, 283, 324, 712, 713
radius, 324
temperature, 280, 324
sunspot cycle, 325
supercooled, 40, 43, 408, 551, 599, 614, 626,
631, 639
drizzle, 1, 599, 669
liquid water, 40
rain, 599
superposition principle, 118
supersaturation, 538
supersonic transport, 397
surface-area concentration, see particle,
surface-area concentration
surface coverage, 411
surface energy, see energy, surface
surface layer, dened, 105
surface roughness length
for energy, 233
for momentum, 231
for trace gas or particle, 667
for water vapor, 233
surface tension, 484, 485, 533
liquid water, 490
containing inorganics, 534
containing organics, 533
surrogate species method, see chemical
mechanisms, condensed
synoptic-scale motions, 8
T
Taylor series
expansion, 128, 173, 174, 175, 178, 188, 198,
201, 421, 422, 423
solution to ODEs from, 4212
Taylors theorem, 175
temperature, 11, 18
absolute, 18
dew point, 45, 46, 59, 602, 606, 607, 701
drop surface, 62831
foliage, 2623
foliage air, 261
frost point, 45, 408
ground, 233, 236, 249, 255, 259, 261, 262, 263
liquid water, 19, 26870
peak stratospheric, 27
potential, see potential temperature
road, 2658
roof, 2658
sea ice, see sea ice, temperature
soil, see soil, temperature
snow, see snow, temperature
sun, see sun, temperature
vertical prole, 1828, 46, 71415
virtual, 33
zonally averaged, 25
temperature inversion, 22, 23, 24, 54, 55, 291,
375, 378, 402, 605
large-scale subsidence, 23
temporally paired peak accuracy, 699
terminal fall speed, see sedimentation, speed
809
Index
terpene, 338, 388
test space, 193
thermal, 22, 24
thermal accommodation coefcient, 532
thermal conductivity, see conductivity, thermal
thermal diffusivity, see diffusion coefcient, for
energy
thermal-infrared radiation, see radiation, infrared,
thermal
thermal speed
air molecule, 18
particle, 509
trace gas molecule, 19, 291, 410, 416, 505,
506, 509, 516, 518, 528, 530, 532, 663
thermal turbulence, see turbulence, thermal
thermodynamic activity, see activity,
thermodynamic
thermodynamic data, 7401
thermodynamic energy equation, 4, 7880, 87,
21314
in a cloud, 610
in terms of potential virtual temperature, 78
thermodynamic equilibrium, see equilibrium
thermo-electric effect, 639
thermometer, liquid-in-glass, 2, 3
thermophoresis, 517
thermophoresis
thermoscope, 2
thermosphere, 19, 28, 57, 282
third body, see M
thunder, 1, 3, 638
time and volume averaging, 6874
time derivative
of chemical concentration, 351
time-derivative conversion
altitude to pressure coordinate, 145
altitude to sigma-altitude coordinate, 161
time interval, 1701, 689
nesting, 691
time required for particle to fall, 665
time-series plots, 699
time-splitting, see operator-splitting
time step, 139, 1701, 689
time-stepping schemes, 181, 2224
AdamsBashforth, 45, 113, 132, 145, 146, 170,
178, 186, 187, 192, 224, 262, 287, 343,
431, 458, 570, 591, 671, 690, 694, 697,
698, 699
CrankNicolson (trapezoidal), see
CrankNicolson scheme
forward Euler, see nite-difference method,
forward Euler
forward-in-time, centered-in-space, 182
Heun, see Heun scheme
implicit, see implicit time-stepping scheme
leapfrog, 1856
Matsuno, see Matsuno scheme
RungeKutta, see RungeKutta method
tire particles, see emission, tire particle
toluene, 357, 386
structure, 392
toluene-hydroxyl radical adduct, 388
topographic data, 692
tornado, 8, 682
Torricelli, Evangelista, 2
total derivative, 63, 77
Cartesian-altitude coordinates, 65
Cartesian-pressure coordinates, 1478
Cartesian-sigma-altitude coordinates, 162
Cartesian-sigma-pressure coordinates,
155
spherical-altitude coordinates, 91
total internal reection, 288, 289
total organic gas (TOG), 338
trace gas, see gas
transfer speed, 667, 675, 676, 679
airsea, 675
transition regime, 507, 509
transmission, 21, 294, 297, 298, 304, 334
transpiration, 16, 255, 261, 262, 264, 268
transverse wave, 118, 119
trapezoidal scheme, see time-stepping schemes,
CrankNicolson
trial function, 193, 195
trial space, 193
triammonium bisulfate (letovicite), 448, 739
trichlorouoromethane, see CFC-11
tridiagonal matrix, 183
trimethylbenzene, 392
trimethylpentane, 392
tropopause, 25, 26, 58, 131, 133, 225, 363, 373,
402, 403, 601
troposphere, 19, 246
truncation error, 173
turbulence, 11, 19, 21, 22, 24, 25, 68, 78, 89,
105, 107, 112, 113, 114, 228, 230, 231,
2357, 241, 244, 245, 247, 251, 252, 254,
271, 273, 274, 477, 519, 521, 610, 661,
664, 666, 667
mechanical, 212, 73
thermal, 22, 73
turbulence parameterization
hybrid, see hybrid turbulence parameterization
TKE, 252
E-
d
, 254
E-
e
, 2524
turbulent ow, 69
810
Index
turbulent ux, 22830
divergence, 70, 82, 105, 591
kinematic, 70, 80, 237
moisture, 228
momentum, 106, 228, 237, 245
sensible heat, 79, 228, 230, 239, 258
vertical
moisture, 228, 230, 239, 261, 262, 667
momentum, 228, 229, 667
sensible-heat, 228, 230, 239, 260, 271
turbulent kinetic energy (TKE), 235, 250, 252,
253, 254, 523
dissipation rate, 236, 253, 511
turbulent Prandtl number, 242
turbulent wake, 105
two-stream method, 328, 331
Tyndall absorption, see absorption
Tyndall scattering, see scattering
U
ultraviolet absorption, 68, 105, 233, 235, 236,
251, 256, 261, 275, 302, 471, 473, 612,
666
ultraviolet radiation, see radiation, ultraviolet
undissociated species, 448
unit conversions, 70913
distance, 709
energy, 710
force, 710
mass, 709
power, 710
pressure, 710
speed, 710
temperature, 710
volume, 709
unit vector
Cartesian coordinate, 61, 64, 83, 84, 86, 91,
96, 104, 107, 115
cylindrical coordinate, 116
spherical coordinate, 84
universal gas constant, 29
universal gravitational constant, 97, 711
universal transverse Mercator, see coordinates,
universal transverse Mercator
unpaired-in-time, paired-in-space error, 697
unpaired-in-time, unpaired-in-space error, 698
unstable stratication, see stability, buoyant
unsteady ow, 69
upslope fog, see fog, upslope
uptake coefcient, see reaction probability
urban smog, see smog, urban
V
vacuum, 2, 275, 277, 285, 310
valence, 26, 3369
principal, 337
valence shell, 336
van der Waals forces, 51214
collision correction, 513
interaction potential, 513
vant Hoff equation, 561
vant Hoff factor, 535
vapor pressure, saturation, 40
trace gas, 533
water over ice, 43, 549
water over liquid, 41
variance, 698, 707
biased, 698
unbiased, 698
vector, dened, 61, 96
vector processor, 686, 687
vectorization, 436
vegetation, 255, 259, 265, 479
velocity, see wind velocity
ventilation factor
ice crystal, 550
liquid drop, 520, 531
thermal, 520, 532
vermiculite, 474
Virgil, 1
virtual boundary conditions, see model, boundary
conditions, virtual
virtual potential temperature, see potential
temperature, virtual
virtual temperature, see temperature, virtual
viruses, 481
viscosity
eddy, see diffusion, eddy
molecular, 101
air, 102: dynamic, 102, 104, 232, 506, 519,
533, 661, 663; kinematic, 102, 232, 236,
506, 529, 531
gas, 101
liquid, 101
viscous force, see forces, viscous
viscous stress, see stress, shearing
visibility, 11, 31316
meteorological range, 299, 300, 301, 314, 315,
334, see also Koschmieder equation
prevailing, 313
visual range, 313
visible light, see radiation, visible
volatile organic carbon (VOC), 338
volcanos, 37, 338, 341, 373, 458, 477, 492
emission, see emission, volcanic
volume, single-particle, see particle, volume, single
particle
volume average mixing rule, 309
volume concentration, see particle, volume
concentration
volume-equivalent radius, 515, 516, 523
volume fraction of particle, 309, 334, 335, 498,
503, 621, 635
volume mixing ratio, see mixing ratio, volume
811
Index
von K arm an constant, 233, 240, 252, 667, 711
von Neumann, John, 5
vorticity, 131
absolute, 131, 214
Earths, 131
geostrophic potential, 134
potential, 131
relative, 131, 214, 215
vorticity equation, 214
W
warm front, 601
washout
aerosol, see coagulation, aerosol-hydrometeor
gas, 10, 6367
water
liquid, 77
density, see density, liquid water
emissivity, see emissivity, liquid water
temperature, see temperature, liquid water
solid (ice), 77
vapor, 7, 10, 11, 12, 14, 16, 20, 2558, 77, 78,
140, 192, 212, 230, 235, 236, 239, 249,
255, 260, 261, 271, 274, 275, 2914, 296,
297, 311, 343, 346, 359, 398, 408, 477,
48791, 51729, 536, 539, 541, 548, 551,
559, 563, 570, 580, 597619, 628, 629,
633, 634, 643, 667, 668, 672, 679, 680,
693, 711, 712, 73851
absorption, 293
mass density, see density, water vapor
mass mixing ratio, see mixing ratio, mass
prole in surface layer, 249
scale, 246
vertical prole of, 24750
water activity, 571
water density, see density, liquid water
water equation, 223, 5704, 577, 620, see also
hydration
water transport equation (soil), 257, 266, 267
water vapor, see water, vapor
wavefront, 286, 287
wavelength, 21, 119, 121, 127, 128, 129, 275,
276, 277, 279, 280, 281, 282, 283, 285,
286, 289, 290, 292, 293, 294, 295, 296,
297, 298, 299, 300, 301, 302, 303, 304,
305, 306, 307, 308, 311, 315, 316, 321,
324, 325, 326, 330, 331, 333, 334, 335,
351, 715, 716
wavenumber, 119, 121, 125, 126, 127, 198, 275,
276, 296, 297, 298
vector, 119
waves, 82, 11835, 136
acoustic (sound), 11929, 13840, 144, 151,
163, 166, 181, 205
internal, 127
acoustic-gravity, 1247
external
amplitude, 119, 121, 123, 126, 127, 134, 205
dispersion relationship, 119, 124, 134
dispersive, 120, 121, 126, 137
displacement, 57, 118, 119, 120, 121, 282
electromagnetic, 21, 37, 119, 120, 273, see also
radiation
envelope, 118, 120, 122
external, 126
frequency of oscillation (angular frequency),
119, 124, 275
gravity (buoyancy), 57, 1247
internal, 125, 126, 127
inertia
low-frequency, 125, 126, 127, 129
group, 118
group speed, 120
group velocity, 120
scalar, 120
internal, 125
Lamb, 1247
external, 127
inertia, 12730
light, see waves, electromagnetic
longitudinal, 119
mountain lee, 127
nondispersive, 120, 121, 124, 127, 137
phase speed (wave speed), 119
Rossby (planetary), 8, 118, 131
forced topographic, 131
freely propagating, 131, 134
transverse, 118, 119
water, 119, 121
weather, dened, 6, 7
forecasting history, 15, 6
weathering
chemical, 474
physical, 473
WegenerBergeronFindeisen process, see
Bergeron process
wetness functions, 258
wetted surface, 490
Wiens law, 273, 282, 334
wilting point, 261
wind
around highs and lows, 11718
geostrophic, see geostrophic wind
gradient, see gradient wind
gust, 105
surface layer, 82, 11314, 11718
westerly, 62
wind direction, 62
wind scalar velocity, 61
angular, 116
geostrophic, 112
horizontal, 85
radial, 116
812
Index
tangential, 116
time- and volume-averaged, 106
vertical
in a cloud, 611
sigma-pressure coordinate, 155,
210
zonally averaged, 62
wind shear, see shear
wind speed, 11, 62
horizontal, 62
logarithmic prole of, 24750, 271
wind velocity, 612
geostrophic, 113
time- and volume-averaged, 106
vector
absolute, 88
angular, 116
horizontal, 85
local, 85
radial, 116
tangential, 116
windmill, 1
work, see energy
X
xenon, 16
xylem, 255, 668
xylene, 357
structure, 392
Y
Youngs relation, 490
Z
Zeldovich nonequilibrium factor, 487, 489
zenith angle, 278, 279, 316, 317, 318, 319
solar, 316, 31719
zonal wind, 62
ZSR method, 571, 576
data for, 7489
Zuni Indians, 1
813

Fundamental of Atmospheric Modeling - Jacobson

Hochgeladen von

Dokumentinformationen

Originalbeschreibung:

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Fundamental of Atmospheric Modeling - Jacobson

Hochgeladen von

Copyright:

Verfügbare Formate

This page intentionally left blank

Fundamentals of Atmospheric Modeling

). In the second, they reduce the nitrite ion to molecular nitrogen

C). Soil has a lower

175 - - 245 nm (2.10)

> 310 nm (2.14)

is the universal gas constant (0.0831451 m

has a value of 2.8704 m

are given in Appendix A.

. The resulting temperature is the virtual

C. This is the temperature at

C, estimate the partial pressure of water, the saturation

is the change in internal

is the work (J) done by or on the parcel. When

- 0, energy is transferred to the environment from the

> 0, work is done by the parcel.

- 0, work is done on the parcel. Equation (2.69) states that if energy

) in both hemispheres, winds originate from the west. In the

= (iu +jv +kw)

k = 1. Cross terms are zero (i

a). The dot product of a gradient operator

(iu +jv +kw) =

v is the divergence of velocity. The equations are also advective forms of

v = 0 into (3.20) give

is the instantaneous perturbation concentration. A precise

= 0), which is the Reynolds assumption.

are perturbation scalar velocities, and

is a perturbation velocity. Advection is the mean horizontal velocity. Thus, u

) are parameterized. Some

with respect to time give

. Meridians extend westward to 180

(180W) longitude and

(180E) longitude. The Equator is dened to have latitude 0

(90S) latitude to +90

(90N) latitude. On the spherical-

v in spherical coordinates gives

(iu +jv +kw) (4.24)

does not change in time at a given location, i

N latitude assuming Lx = 500 km, Ly = 500 km, and Lz = 10 km for

to the right of the direction of motion, and in

to the left of the direction of motion. The

N, only the vertical component

2Owcos may be removed from (4.32). Because the vertical component of

is the true gravitational acceleration. The vectors i

and g vary with altitude in the Earths atmosphere. Equation (4.40)

1012 1008 hPa

1013 1000 hPa

to the right of the Coriolis force.

counterclockwise in comparison with winds aloft

clockwise in comparison with winds aloft

= i sin +j cos k = k (4.85)

are the perturbation components of u

k (sound waves), where c

. Thus, in a nondispersive medium,

Thus, individual deep-water waves in a group move twice as fast as the

, and eliminating products of perturba-

, (4.115) can be solved analytically to

k {mountain lee waves

when H = 8.72 km, c

, respectively, and the equa-

is reasonable for midlatitudes,

give the dispersion relationship for freely propagating Rossby waves

. Calculate dx at the north and south boundaries of